## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Agent runtime settings can bind organization secrets to an adapter
environment
> - Paperclip redacts plain environment values when it returns a saved
agent to the UI
> - A saved-agent test sent the redacted `CODEX_HOME` value back to the
server
> - Codex ACP also received the API key without an ACP API-key
authentication request
> - This pull request restores saved environment values for tests and
selects API-key authentication for Codex ACP runs
> - The benefit is that Codex agents can test and run with an
organization-scoped OpenAI API key
## Linked Issues or Issue Description
**What happened?**
Testing a saved Codex agent sent `***REDACTED***` as `CODEX_HOME`.
Secret normalization rejected that placeholder. Remote Codex ACP runs
received `OPENAI_API_KEY`, but session creation stopped with
`Authentication required`.
**Expected behavior**
Paperclip must use the saved `CODEX_HOME` value when it tests an
existing agent. Codex ACP must select API-key authentication when
`OPENAI_API_KEY` is available.
**Steps to reproduce**
1. Create an organization-scoped secret named `OPENAI_API_KEY`.
2. Give a Codex agent access to the secret.
3. Save the agent runtime settings.
4. Test the saved agent again.
5. Run the agent in a remote sandbox through ACP.
**Paperclip version or commit**
Reproduced on master before commit
`68c17709d7c051a804a416263e2e08920f1dfcb1`.
**Deployment mode**
Self-hosted server with a remote sandbox environment.
**Installation method**
Built from source.
**Agent adapter(s) involved**
Codex.
## What Changed
- Send the saved agent ID with adapter environment tests.
- Restore redacted plain environment values from the saved agent before
test-time secret resolution.
- Select the Codex ACP `api-key` authentication method when
`OPENAI_API_KEY` is present.
- Add focused regression coverage for saved-agent tests and remote ACP
launch configuration.
## Verification
- `pnpm --filter @paperclipai/adapter-utils exec vitest run
src/acpx-engine/execute.test.ts`
- `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/agent-adapter-validation-routes.test.ts`
- `pnpm --filter @paperclipai/ui exec vitest run
src/lib/test-agent-setup.test.ts`
- `pnpm -r typecheck`
- `pnpm test:run`
- `pnpm build`
- `git diff --check`
## Risks
- Low risk. The test route reads saved configuration only when the
request supplies a compatible agent ID and the caller can update that
agent.
- The Codex ACP change applies only when `OPENAI_API_KEY` exists and no
explicit `DEFAULT_AUTH_REQUEST` exists.
- There are no schema migrations or telemetry changes.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex with `gpt-5`. The context-window size is not exposed in
this runtime. The model used reasoning, repository search, file editing,
command execution, and test execution.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
The Claude card's button only moved to Connecting once the login was stored - a submit, a status poll and a completion read after the paste - so for about a second the customer had done their part and the button still read Waiting for code. The panel now reports the submit as it starts (onCodeSubmitted) and the step shows Connecting from that moment.
That could not simply move the phase earlier: the two-second hold started when Connecting did, so it would have hired whether or not a credential existed. The hire now waits for both the stored login and two seconds of Connecting counted from the paste. onSubmitFailed gives the button back when a submitted code does not become a stored login, the field locks while a code is out, and Cmd+Enter no longer hires mid-connect.
Reports that land after Back are ignored. The panel stays mounted through Back's exit, so a late failure reopened the card being left, and a late success hired a customer who had backed away. The second predates this change; its test fails the same way against master.
The authorization code shows as dots. The OpenAI card is untouched.
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - The `codex_local` adapter signs agents in to OpenAI, and a company
keeps one default Codex identity in its shared company home
> - A login with a DIFFERENT account than the company default is
deliberately kept out of the shared home — one agent's sign-in must not
switch every unbound agent's credentials — but that left the
cross-account login inert: nothing connected the agent the operator was
configuring to the credential the login stored
> - The stored credential and its company secret already exist; only the
last mile — an agent actually using them — was missing
> - This pull request reports a non-secret binding claim on the
authenticated login and lets the agent page bind that one agent's
`CODEX_HOME` to the account's secret, exactly and only when the
identities differ
> - The benefit is that multi-account Codex becomes one click on the
agent that needs it, with company-wide identity untouched
## Linked Issues or Issue Description
**What happened?**
On an agent's detail page, "Sign in with Codex" using a different OpenAI
account than the company default succeeds but changes nothing for that
agent. The credential lands in the per-identity store and a company
secret names it, but the agent keeps using the company default. The Test
keeps reporting that authentication is needed, and no repeat login
helps.
**Expected behavior**
When the operator deliberately signs an agent's page in with a different
account, that agent starts using that account. Agents that were not part
of the action keep the company default. A same-account login keeps
working through the shared company home with no per-agent pinning.
**Steps to reproduce**
1. Configure a company whose Codex home holds account A.
2. Open a `codex_local` agent's detail page with a sandbox environment
and complete "Sign in with Codex" using account B.
3. Press Test. Before this change the agent still resolves account A and
the authentication-needed check returns.
## What Changed
- `packages/adapters/codex-local` — the prerequisite shield:
`isCodexAuthCachePath` recognizes per-identity credential-store entries,
and `seedManagedCodexHome` refuses to symlink, heal, or
API-key-overwrite an entry's `auth.json`. The seeding pass runs before
every probe and execute; without the shield, an agent bound to an entry
would have its stored login silently swapped for the host credential.
Static shared config files still copy in. Rotation already survives
binding: the sandbox copy-back writes rotated credentials into the
identity-keyed store slot.
- `server` — the promotion records whether the company default home
ended on a different account than the login (any read failure degrades
to `false`, so the client can never be told to bind wrongly). After the
terminal commit, the routes layer remembers a non-secret claim — the
opaque account-home secret id plus that verdict — in a bounded in-memory
map, and merges it into the owner read of an `authenticated`
`codex_local` session. A restart drops the claim; the panel then shows
plain success.
- `packages/shared` — `CodexAccountBindingClaim` on the owner session
response. It carries no account identifier and no credential byte.
- `ui` — the login panel reports the claim upward once. The edit-mode
form binds the agent's `CODEX_HOME` to the secret and saves in one step,
only when `companyIdentityDiffers` is true. Same-account logins bind
nothing on purpose: the company-home refresh already carried them, and
an unbound agent keeps following the company default across rotations.
Create mode is unchanged.
## Verification
- Adapter suite: 381 passed, 1 skipped (includes the new store-entry
shield tests and the path-predicate cases).
- Server suites (8 files): 130 passed, 15 skipped — including two new
route tests that drive a login to `authenticated` and assert the claim
with both identity verdicts.
- UI render suite: 85 passed — including a panel test that the claim is
reported upward exactly once.
- `tsc --noEmit` clean in `packages/shared`, the adapter package, and
`ui`; `server` clean for the touched file.
## Risks
- The bind changes one agent's configuration through the normal
agent-update patch, initiated by the operator's own login on that
agent's page. The failure direction of every fallback is "offer
nothing": a missing claim, a restart, or an unreadable company home all
degrade to no bind.
- The seed shield narrows what the seeding pass may touch; homes outside
the credential store behave exactly as before, covered by the existing
seed tests.
- Builds on the sign-in credential-resolution fix (#13064), now merged;
this branch is rebased onto master and the diff contains only the
binding feature. Supersedes #13066, which GitHub auto-closed when its
stacked base branch was deleted on merge.
## Model Used
Claude (Anthropic) — Claude Fable 5 (`claude-fable-5`), extended
thinking, agentic tool use in Claude Code (terminal).
**Related PRs (searched; no duplicates found):** #12740, #12082, and
#9621 touch adjacent Codex credential sync paths; #8495 is the standing
hardening effort for probe auth seeding.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes (no
standalone docs cover this flow; the behavioral contracts are documented
in-line at each changed site)
- [x] I have considered and documented any risks above
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent adapters select a provider, a model, and a runtime.
> - Runner conversion rejected existing Claude agents. The model list
mixed providers.
> - The native Claude runner rejected custom models and could not launch
on macOS.
> - This pull request fixes conversion, model selection, and verified
macOS execution.
> - It also groups configuration fields consistently across adapters and
opens artifact images in the task gallery.
> - Operators can change an agent configuration and run the selected
model on their Mac.
## Linked Issues or Issue Description
**What happened?**
Converting an existing Claude agent to Paperclip Runner failed with a
Codex-only restriction. ACPX Claude showed unrelated models and required
`claude-sonnet-5`. Its native runtime rejected macOS. Configuration
mixed common model settings with process controls. Artifact cards
labeled “Open gallery” navigated to attachment URLs instead of opening
the task gallery.
**Expected behavior**
Conversion keeps agent identity and compatible settings. ACPX Claude
uses the normal Claude catalog and accepts typed model IDs. Codex uses
the native runner. The verified Claude runtime can launch on macOS ARM64
and x64. Common configuration sections place the same fields together
across adapters. Artifact images open in the shared task gallery with
navigation and downloads.
**Steps to reproduce**
1. Open the configuration of an existing Claude agent.
2. Convert it to Paperclip Runner.
3. Select ACPX Claude and a different catalog model or a typed model ID.
4. Save the agent and run a disposable task on macOS.
5. Inspect configuration and advanced run-policy controls across
adapters.
**Paperclip version or commit**
The bugs were reproduced on `165ca56a22adb60e5fda56045442d9c8498116a8`.
This branch was rebased onto `7ed122911`.
**Deployment mode**
Built from source. Local test-drive instance on macOS ARM64 with an
isolated database.
Related work: #11798 addresses unsupported ACP session options in the
existing adapter path. #13048 addresses working-folder preservation.
This change fixes native runner configuration and launch behavior.
## What Changed
- Remove the Codex-only conversion restriction. Preserve agent identity,
instructions, directories, credentials, and compatible model settings.
Reset incompatible sessions while retaining history.
- Show ACPX Claude and native Codex as distinct provider choices. Remove
ACPX Codex from advertised configuration. Normalize legacy
configurations before fresh runs without rewriting historical run
descriptors.
- Select model catalogs and cache entries by provider. Support refresh
and typed model IDs. Pass exact Claude IDs through session creation,
model changes, and recovery.
- Add verified macOS ARM64 and x64 Claude SDK snapshots. Bound
executable allocation and total snapshot size. Preserve package checks,
dependency isolation, process ownership, cancellation, and Linux
descriptor loading.
- Probe local runtime readiness. Report remote platform checks as
incomplete until the remote runner verifies its runtime.
- Surface actual model rejection and allow correction and retry.
- Repair missing ACPX goal-capability helpers exposed by the post-rebase
live test. Persist and restore the optional capability without breaking
session startup.
- Put Agent identity first and intentionally remove the Capabilities
editor, as requested. This is removal of UI editing, not relocation:
preserve existing capability metadata and API compatibility without
adding another editor. Use the themed select for configurable permission
modes, with normal text instead of monospace.
- Put model and provider under Adapter. Give environment variables their
own section. Fold command and arguments under Configuration. Fold
lifecycle, timeout, and interrupt grace under Advanced Run Policy. Hide
single-option permission controls.
- Open image and video artifact cards in the existing task gallery,
including cards in the artifacts panel. Chat attachment images use the
same gallery. Preserve standalone media previews and download links.
## Verification
- Rebased focused UI/API/database suites: 293 tests passed.
- Rebased native runtime and ACPX suites: 242 passed, 7 skipped.
- Repository typecheck, build, and token gates passed for the runner
changes. Gallery follow-up UI typecheck, build, and token gates also
passed.
- Follow-up UI suites passed (86 tests), packaging checks passed (14
tests), and the final focused runtime suites passed (126 passed, 7
skipped).
- Linux container isolation and lifecycle fixtures passed before rebase
(57 passed, 2 skipped). Rust ACPX provider-session tests passed after
rebase (8 tests).
- Browser tests completed actual Claude and native Codex tasks on macOS
ARM64. They covered conversion, catalog refresh, a non-default catalog
model, a typed `haiku` ID, save/reload, cancel, follow-up session
continuity, invalid-model errors, and recovery.
- Final-revision live tests completed a typed Claude task, a follow-up
with the same provider session, and a native Codex task on macOS ARM64.
- Browser tests confirmed the moved interrupt-grace field saves and
survives reload. Cross-adapter tests cover Claude, Codex, Gemini,
process, gateway, and schema forms.
- Full local run: 7,080 passed, 30 skipped, and two timeouts. Both
timeout suites passed on isolated rerun (84 tests); the failures were
the plugin login-worker exit diagnostic and the runner real-server
vertical slice.
- Final follow-up checks: 50 registry tests and 45 snapshot/installation
tests passed (6 platform-specific skips). Oversized executable rejection
is covered before allocation or reading; unsupported-platform tests
invoke the real installation probe.
- Runner head `ddb5101c483a297f74875ab96b3c66035b002d50`: all CI gates
green, including full runner verification, repository build, typecheck,
general/serialized server suites, browser tests, and canary dry run. [CI
run](https://github.com/paperclipai/paperclip/actions/runs/34286178670).
- Greptile: 5/5 on that runner head. All four review threads resolved.
Superagent, Socket, and Snyk checks green.
- After snapshot hardening, another real Claude task completed on this
Mac using the rebuilt runtime.
- Gallery follow-up: 148 focused tests passed, covering artifact
selection, shared attachment collections, deduplication, image/video
cards, standalone previews, downloads, and closing. Live browser
verification completed on the settings follow-up: artifact selection,
6-image pagination with wrapping, download action, and closing all
stayed on the same task URL. All checks passed on gallery head
`96136da58ff195bf6ca00b281eb3022ad12d7bd8`: [CI
run](https://github.com/paperclipai/paperclip/actions/runs/34287987536).
Greptile returned 5/5 on that exact head with no unresolved threads.
- Final settings polish: 96 focused tests, UI typecheck/build, and token
gates passed. A real browser walkthrough verified readable permission
options, identity placement, Capabilities removal, and permission
save/reload. Original test-agent permission mode restored. All 31 checks
passed on final head `e46540d6bf32bfb0566dca16b2f4a75ba437618c`: [CI
run](https://github.com/paperclipai/paperclip/actions/runs/34292797886).
Greptile returned 5/5 with no unresolved threads.
## Risks
- Capabilities intentionally has no editable UI field after this change.
Existing values remain readable and API-compatible; removing the field
does not erase stored metadata.
- macOS launch now copies verified package files into private snapshots.
The implementation must retain isolation and clean up snapshots on exit.
- Runtime provider or model changes reset the current session.
Historical runs remain available.
- The macOS x64 SDK executable digest was verified, but a live Intel Mac
run was not available. Linux verification used container fixtures, not a
real Claude task.
- Remote environment tests report a warning when only the platform has
been checked. They do not claim package readiness from the server host.
## Model Used
OpenAI Codex, based on GPT-6. The exact served model identifier and
context-window limit are not exposed in this session. Used reasoning,
repository inspection, code execution, Rust and TypeScript tests, and
browser automation.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass (focused suites and both
timeout suites on rerun; full-run counts above)
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Resolve unset Claude models to Opus 5 across CLI and ACP execution, preserve explicit and provider-specific overrides, and show the default in agent configuration.
Verified 212 focused tests after merging master, UI typecheck and token gates, and all CI checks. Greptile reviewed the final head at 5/5.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Operators create agents and configure their runtimes in the board
UI.
> - The old creation flow presents several choices and a large form
before an agent can start.
> - The existing onboarding controls already provide clear provider
connection steps.
> - This pull request uses those controls in a new-agent wizard and
organizes the full configuration pages.
> - Operators can connect, test, save, and assign a first task while
keeping the existing configuration tools.
## Linked Issues or Issue Description
Related: #10974. That earlier open PR also reorganizes agent
configuration. This PR follows the reviewed Storybook designs for agent
creation and the current configuration tabs.
**What existing behavior does this improve?**
Agent creation, provider connection, runtime tests, and full agent
configuration.
**Current behavior**
The creation dialog leads to a large manual configuration form. Provider
login controls differ from onboarding. Environment variables and secret
access appear in separate places.
**Proposed behavior**
Choose a name and adapter. Connect Claude or Codex through the existing
onboarding controls. Configure and test the runtime, save the agent, and
open a task dialog with that agent assigned. Use the same design on the
existing configuration tabs.
**Reason and benefit**
The first setup asks for fewer decisions. The full editor keeps
instructions, skills, runtime controls, secret access, permissions,
keys, and revisions available in clear sections.
**Breaking changes**
The board creation and configuration layouts change. The
test-environment API adds an optional, allowlisted `testCredentials`
field for one-shot probes. Database contracts stay the same. Native ACPX
tests now reject unsupported local platforms before a CLI login can mask
the runtime restriction.
## What Changed
- Added a new-agent wizard with numbered steps, adapter branding,
provider connections, editable model choices, runtime tests, and
confirmation.
- Added Codex app-server, Claude ACPX, and OpenCode runner choices.
- Stored API credentials through existing secret APIs and persisted
references in agent configuration. New setup keys are isolated from
credentials used by existing agents.
- Preserved external-agent invitations beside the wizard, including
optional messages, one-time prompts, and clipboard fallback.
- Added OpenRouter provider and secret bindings for Pi and OpenCode.
- Added adapter-specific prerequisite fields for Cursor, Gemini, Kimi,
and Hermes. Cursor Cloud keys are saved as new organization secrets.
- Fixed Cursor Cloud repository field mapping, omitted empty remote
environment values, and added useful model and repository error
messages.
- Preserved complete MCP assignments when multiple valid profiles
contain more than 250 tools in total. Generated profiles retain exact
tool selectors.
- Added service branding and deployment-aware adapter choices. Cloud
setup offers Claude, Codex, and OpenCode; local native runners require
the experimental setting.
- Made the agent list responsive at intermediate widths.
- Applied the reviewed design to the real agent configuration pages.
Kept the instruction editor, skills, and existing mutations.
- Combined secret access and environment variables under one Save and
Discard action.
- Added interactive Storybook screens for setup, configuration,
confirmation, authentication, and test results.
- Fixed Pi provider-error parsing and thinking-effort persistence.
Native ACPX validates Linux x64 on the actual local, SSH, or sandbox
target.
- Redacted the complete transient probe-credential field from HTTP error
logs, including rejected provider names.
## Verification
- Current head `df0292fe6` has a fresh Greptile 5/5 review with no
unresolved findings. All 31 executed CI checks passed, including the
aggregate verification gate and all browser E2E shards. Storybook visual
regression is skipped by its workflow; the local Storybook build passed.
- Browser tests completed real assigned tasks with direct Codex, Claude,
OpenCode, Pi, and native Codex.
- Verified external-agent invitation generation and automatic prompt
copying in the live browser.
- Pi and OpenCode used an existing OpenRouter secret. Browser checks
covered save and reload, instruction edits, skill selection,
environment-variable Save and Discard, and assigned task creation.
- Invalid Claude API credentials remained on the connection step with an
error. A live Pi/OpenRouter invalid-key probe returned a provider
failure and left the user-secret inventory unchanged (zero entries
before and after).
- Full workspace typecheck and build passed after rebasing onto current
master. After review fixes, server and UI typechecks, token gates, and
the full build passed again. Storybook built successfully.
- All 5,542 local UI tests passed. The Cursor Cloud and Pi adapter
regressions passed all 24 tests. Review regressions passed 69 server
tests and all 18 agent-list tests.
- The local full test command ran 6,971 general server tests
successfully. Editing review fixes during that long run caused nine
tests to use stale modules; fresh isolated runs passed. An unrelated
embedded-Postgres fixture hit the host shared-memory limit; its 15
affected tests passed when the fixture groups ran separately.
- Local workspace groups passed after rerunning 18 CLI tests
sequentially to avoid host database limits and parallel-load timeouts.
The local full command stopped at the general server phase, so
serialized server verification comes from the five passing CI shards.
- Browser testing at 390px confirmed that the agent action menu opens
and the page has no horizontal overflow. CI browser E2E shards passed.
- Review the `Onboarding / New agent` and `Agents / Configuration
refresh` Storybook groups. In the real app, create an agent, run its
connection test, save it, assign a task, and reload its configuration.
## Risks
- This changes the main agent setup and configuration UI. Regression
tests cover routing, persistence, secret bindings, and form actions.
- Native Claude ACPX requires Linux x64. Direct Claude works on macOS.
Remote checks execute a bounded platform probe and reject unsupported or
unverified targets.
- A native OpenCode task reached the provider context limit because of
its tool payload. Its provider connection test passed. Direct OpenCode
completed a task. This existing native execution limit is not fixed
here.
- Claude and Codex connection keys use the existing user-secret store.
Other runtime setup keys use distinct organization secrets. Existing
credentials are never rotated. Probes do not store entered keys. Failed
agent creation removes newly staged credentials.
- Cursor Cloud has not completed a live task. Its authenticated account
still needs GitHub repository access. The live run passed MCP
provisioning, remote environment validation, and explicit Auto model
selection before the repository prerequisite blocked execution.
- Generated runtime MCP profiles can exceed the public profile-edit
request limit. They still contain exact catalog selectors and preserve
permission boundaries.
- No database migration, dependency, lockfile, or workflow changes are
included.
## Model Used
OpenAI Codex, based on GPT-6, with reasoning, repository tools, shell
execution, and browser automation. The runtime did not expose the exact
model ID or context window size.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
The card carried a Cancel beside its instruction, directly above the step's own Back. Two ways out of one screen is one too many.
Removing it was only safe if returning genuinely resumed the running session, and writing the test for that disproved the premise: coming back started a second login rather than adopting the first, which the server's per-owner lease rejects. Cancel was not duplicative — it was the only way out of a login the customer could no longer reach.
The cause was the cache. The active-session read is a useQuery with no staleTime, so on the second mount it answered instantly from the first mount's cached null: isFetched was true, the resume effect concluded there was nothing to adopt, and autoStart began a new login while the refetch was still in flight. Both panels did it, and it was pre-existing on master. Both reads now refuse the cache, so the resume works and the removal is safe because of it.
Removing the button also stranded the chain behind it: handleCancel in both panels, the onCancel prop on AdapterLoginPanel, and the wizard's wiring. The panel chrome's own Cancel is untouched.
An abandoned login is collected on a five-minute timer (DEVICE_LOGIN_TIMEOUT_MS) with the reaper as the restart-safe backstop, and the lease is keyed on adapterType, so a source switch does not contend. Both covered by tests.
The Claude card cleared its field the instant the paste was submitted, so the code appeared and vanished in the same breath and the seconds that followed had no feedback at all - it read as the paste having been dropped. The code stays now; the field is disabled and the step's button carries the status. Scoped to the onboarding chrome: the panel in the agent form stays open long after its login and has its own status area, so clearing is still right there.
The OpenAI card's auto-copy took its latch before the attempt rather than on success, so the first refusal was permanent - and the first attempt is the one most likely to be refused, since it fires while the card is still arriving. It also ran with an empty code on the paths where the row renders before the server's one-time prompt has a value, which wrote nothing and then said Copied! over whatever the clipboard already held. It now waits for the document to have focus, retries when focus returns, guards against a second attempt while one is in flight, and only latches once a write has actually landed.
The arc's question is a quarter larger, 36px to 45px, as a token with its own leading. paperclipai/paperclip-cloud#400 moves the signup wizard's matching heading by the same quarter - the two flows are pinned to each other on purpose.
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent authentication uses server sessions, plugin workers, and
browser login panels.
> - A page reload loses an active login session, and one worker permits
only one login terminal.
> - These limits cause lost work and prevent two owners from logging in
through one worker.
> - This pull request lets the browser resume active sessions and lets
workers serve concurrent login terminals.
> - The benefit is reliable login recovery with a bounded process-wide
route limit.
## Linked Issues or Issue Description
**What existing behavior does this improve?**
It improves agent credential login recovery and concurrent login
terminal handling.
**Subsystem affected**
Cross-cutting (multiple of the above).
**Current behavior**
A page reload loses the active login session. A shared plugin worker
rejects a second login terminal.
**Proposed behavior**
The browser reads and resumes the owner's active session. A worker
supports multiple login terminal routes under a process-wide ceiling.
**Reason and benefit**
Owners keep login progress after a reload. Two owners can log in through
one worker without removing the route limit.
**Breaking changes**
None. The change adds owner-scoped read routes and changes login
terminal concurrency.
## What Changed
- Replace the single worker login route with maps keyed by host route
and worker session identifiers.
- Add a process-wide login route ceiling and release each reserved slot
on every exit path.
- Add owner-scoped active-session reads with consistent negative
responses and private cache control.
- Keep the device-login prompt while the session has an active public
status.
- Add a durable setup-token cancel fallback for a lost in-memory
session.
- Resume active sessions when the agent configuration or onboarding
panel mounts.
- Remove routine unmount cancellation and keep explicit Cancel behavior.
## Verification
- `pnpm --filter @paperclip/server test` — server route, service, and
plugin-worker-manager suites.
- `pnpm --filter @paperclip/plugin-sdk test` — worker RPC host suite.
- `cd ui && npx vitest run
src/components/AgentConfigForm.render.test.tsx
src/components/OnboardingWizard.test.tsx`.
- `cd ui && npx tsc -b`.
- `tests/e2e/onboarding.spec.ts` — reload during login.
- CI must pass on this pull request.
## Risks
The change affects agent authentication and the sandbox-to-host
boundary. Route cleanup must release every reserved slot. Owner checks
must prevent cross-owner session access. Tests cover route cleanup,
owner scope, reload recovery, and concurrent worker routes.
## Model Used
Codex, OpenAI GPT-5, tool use and code review support. The
implementation author owns the exact model details for the code changes.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have described the issue in-PR with the relevant issue-template
fields
- [x] I have not referenced internal/instance-local Paperclip issues or
links
- [x] My branch name describes the change and contains no internal
Paperclip ticket id
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Picking a source starts the sign-in: the row collapses to the answer, the card opens where the credential link was, and the footer button walks Sign in -> Waiting for code -> Connecting before the step advances. Back unwinds it a beat at a time.
Nothing mounts to change layout - the card and the link are always rendered and their heights animate, with inert holding the a11y line - because a mount changes the page in one frame and no easing can smooth a step already taken.
Review fixes in the same branch: the displayed-code panel now reports its prompt upward (the OpenAI path could not leave the loading beat without it), the two-second hold is a cancellable beat rather than a dropped timer, unwinding a sequence that never opened a card no longer starts a login to cancel it, and the key field regains focus-on-open.
## Thinking Path
> - Paperclip is the open source app that people use to manage AI agents
for work.
> - The Codex local adapter supplies model metadata to the server and
the user interface.
> - OpenAI now lists `gpt-6-astra` as a supported Codex model.
> - Paperclip did not list this model or its model-specific controls.
> - This pull request adds the model through the existing adapter
metadata path.
> - The benefit is that agents and task overrides can use the exact
model ID and supported controls.
## Linked Issues or Issue Description
**Subsystem affected**
`packages/adapters` and `ui`
**Problem or motivation**
Paperclip does not expose `gpt-6-astra` in Codex model selectors.
Operators cannot select and save the model through the normal agent and
task forms.
**Proposed solution**
Register the exact model ID in the Codex local adapter. Use the adapter
as the source for the model-specific reasoning options. Preserve the
current default model. Forward the saved model, reasoning effort, and
fast-mode controls through both Codex execution lanes.
**Alternatives considered**
A user-interface-only model list would duplicate adapter metadata. A
model alias would not match the official model ID. Both options were
rejected.
**Roadmap alignment**
This is a small adapter compatibility update. It does not duplicate a
planned item in `ROADMAP.md`.
## What Changed
- Added `gpt-6-astra` to the Codex local adapter model registry and
fast-mode support list.
- Added the official Astra reasoning efforts: `low`, `medium`, `high`,
`xhigh`, `max`, and `ultra`.
- Used the adapter metadata in agent and task model selectors.
- Preserved supported effort choices when the model changes. Cleared an
effort only when the new model does not support it.
- Added tests for registration, user-interface selection, configuration
persistence, and CLI and ACP forwarding.
## Verification
- `pnpm exec vitest run packages/adapters/codex-local/src/index.test.ts
packages/adapters/codex-local/src/server/acp.test.ts
packages/adapters/codex-local/src/server/codex-args.test.ts
packages/adapters/codex-local/src/ui/build-config.test.ts
ui/src/lib/codex-reasoning-effort.test.ts
ui/src/components/AgentConfigForm.render.test.tsx
ui/src/components/IssueProperties.test.tsx
ui/src/components/NewIssueDialog.test.tsx
ui/src/lib/issue-assignee-overrides.test.ts` passed 245 tests.
- `pnpm -r typecheck` passed.
- `pnpm check:token-gates` passed all four gates across 939 files.
- `pnpm --filter @paperclipai/ui build` passed and supplied isolated
user-interface build proof.
- `pnpm build` passed.
- `pnpm test:run` passed 5,812 tests and failed 24 workspace-runtime
tests in this isolated host. The failures use invalid generated ports
above 65,535, incomplete nested-worktree fixture configuration, or
`/tmp` path aliases. The focused tests for this change all pass. GitHub
CI must pass before review handoff.
- GitHub CI run `33918372718` passed all required checks and the
aggregate verify gate on exact head
`6ac6be2cee0a5996c82bdf674fcb7f46cb4c5fde`.
- Independent engineering review approved the exact remediation head
after 170/170 reviewer tests passed.
- Greptile reported 5/5 with no open review threads on exact head
`6ac6be2cee0a5996c82bdf674fcb7f46cb4c5fde`.
- The model ID and capabilities were checked against the [official
OpenAI Codex model list](https://developers.openai.com/codex/models).
## Risks
- Low risk. The change adds one model and model-specific selector
options. It does not change the default model.
- OpenAI can change model capabilities later. The adapter metadata must
stay aligned with the official Codex metadata.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex with model ID `gpt-5.6-sol`, a 272,000-token context
window, reasoning, tool use, and code execution.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Connect starts the sign-in; the card that appears is the sign-in rather
than an offer of one; success advances to Review rather than reporting
itself. The two logins end in different places and the button says which:
Claude submits a code back here, so it spins on "Connecting"; OpenAI
finishes in another tab, so it stays a still, disabled Next until the
poll lands.
AdapterLoginPanel grows autoStart / onCancel / onConnected / chrome
rather than a second implementation — the session start, both polls, the
server deadline, the one-shot completion read and the unmount release are
the parts onboarding needs unchanged. Every prop is off by default, so
the agent form and the new-agent page render what they did before.
Claude's code auto-submits on the paste, not on every change:
isValidBrowserCode accepts any printable ASCII from one character up, so
a value-driven submit fired on the first keystroke of anyone who typed.
Also orders the OpenAI card and the settings displayed-code panel
code-above-link, with the instruction worded to match, and releases the
displayed-code session on unmount so an abandoned login stops holding the
one-per-owner reservation.
## Thinking Path
> - Paperclip is the control plane for agents that perform work.
> - Paperclip Runner connects durable provider sessions to individual
task runs through PRP.
> - Provider continuity and per-run authority are different lifetimes.
> - The existing implementation mixed those lifetimes and lost event
metadata between provider frames, runnerd, persistence, API
sanitization, and the task thread.
> - That caused failed continuation, missing progress and Plans,
duplicate replies, hidden failures, and unsafe recovery.
> - This repair gives every heartbeat fresh authority, preserves
qualified provider-session continuity, and restores one lossless
presentation path without changing direct adapters.
## Linked Issues or Issue Description
**What happened?**
A second native heartbeat could reuse tickets, leases, command receipts,
sequence state, and run identity from the first heartbeat. Provider
phase and item identity could be lost before the UI read them. Redaction
could corrupt protocol discriminators while still missing malformed
credential tails. The task thread could fold progress into the final
response, hide failures, or show more than one final answer. Native
Codex also exposed approval modes that do not yet have a durable
approval bridge.
**Expected behavior**
Each heartbeat uses a new PRP authority epoch. Codex and OpenCode
preserve exact qualified provider sessions; ACPX emits an explicit
continuity event when its qualified process-replacement policy is used.
Every accepted provider event is presented, classified as internal, or
surfaced as unsupported. The task page shows chronological progress,
reasoning summaries, activity, Plans, interactions, terminal failures,
and exactly one final reply. Direct adapters retain their existing path.
**Steps to reproduce**
1. Enable the unified experimental Paperclip Runner setting.
2. Create a local native Codex, OpenCode, ACPX Claude, or ACPX Codex
agent.
3. Run response, Plan, structured-question/resume, restart,
cancellation, and failure scenarios.
4. Reload the task while active, waiting, failed, and settled.
5. On the old implementation, observe stale run authority, missing
classifications, incomplete output, or duplicated/folded replies.
**Paperclip version or commit**
The repair is based directly on `master` at
`87d05e194b643810d16d20612115acd01d735d43`.
**Deployment mode**
Local development with the embedded database.
Related work: Refs #12616, #12646, #12666, #12685, and #12700.
## What Changed
- Rotates PRP control-plane, outbox, ticket, lease, command, receipt,
and sequence authority for each heartbeat while carrying forward only a
validated provider-session identity.
- Reads `control-plane-state.json`, validates both durable schemas and
lifecycle values, resumes coherent current runs, archives qualified
settled authority, and quarantines malformed or mismatched scoped state
without moving ambiguous live legacy state.
- Preserves Codex provider phase and stable item identities so
commentary remains progress and only `final_answer` becomes final.
- Adds raw OpenCode HTTP/SSE boundary coverage and canonical reasoning
lifecycle mapping.
- Makes ACPX normalization lossless for visible reasoning, tool
lifecycle metadata, stable bounded identities, Plan revisions,
structured requests, failures, and qualified process replacement. Only
the compatible terminal assistant message is promoted as final.
- Applies schema-aware redaction before generic JWT-shaped detection and
scans every diagnostic string leaf. Malformed raw/escaped quoted
credential tails are redacted in both server and durable Rust state.
- Restores snapshot-style chronological task presentation, expandable
tool activity, inline Plan cards, visible waiting/resume/cancel/failure
states, and exactly one final answer.
- Makes `never` the only qualified native Codex permission mode and
rejects unsupported persisted native modes with remediation. OpenCode
and ACPX policies remain intact.
- Keeps the unified experimental Runner setting as the only enablement
flag. Onboarding and direct Codex, Claude, and OpenCode stay on their
legacy execution/finalization paths.
- Adds cross-language goldens, authority/recovery/fault coverage, exact
response/count assertions, and native plus legacy acceptance scenarios.
## Verification
- Pull-request GitHub Actions run Rust formatting/tests, TypeScript
checks, server/UI tests, builds, protocol drift checks, browser E2E, and
security scans.
- A separate workflow-only validation ref is pinned directly on this PR
head and runs the 35-cell paid local matrix: three core scenarios plus
structured-question resume and restart/resume for native Codex, native
OpenCode, ACPX Claude, ACPX Codex, and direct Codex/Claude/OpenCode.
Run: https://github.com/paperclipai/paperclip/actions/runs/33682434315
- Acceptance requires exact single visible replies, monotonic sequences,
matching envelope discriminators, one semantic terminal, one run
terminal, no unresolved interaction, no duplicate mutation, no secret
leakage, provider continuity, and zero native rows for direct adapters.
- Per maintainer direction, tests are running in GitHub Actions rather
than on the slower local host. Only formatters and static diff checks
were run locally.
## Risks
- Recovery from old or partial filesystem state is sensitive. The repair
fails closed, preserves active or unverifiable authority, and
quarantines only state whose scoped ownership is safe to move.
- Provider event formats can change. Closed validators and boundary
goldens turn new or malformed events into visible diagnostics instead of
silent drops.
- Shared task presentation could affect direct adapters. Runtime-fact
gating plus the direct-adapter matrix protect the existing path.
- Managed and remote providers are not qualified here. Shared code
continues to compile and fail safely, but live qualification is
deferred.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex based on GPT-5. The exact deployed snapshot and
context-window size are not exposed to this task. It used agentic
reasoning, repository inspection, code editing, Git, parallel subagents,
and GitHub Actions.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change and contains no internal
Paperclip ticket id
- [ ] I have run tests locally and they pass (intentionally deferred to
GitHub Actions)
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented risks above
- [ ] All Paperclip CI gates are green
- [ ] The paid local-provider matrix is green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
Takes the four tenant onboarding steps to the design, and makes the model
choice explicit.
**Nothing is preselected on the connect step.** It arrived with a source
already chosen, which made the row read as a confirmation rather than a
question and let a customer pass the step having touched none of it. The
step now opens unanswered and cannot advance until a source is selected in
the visible row.
Two defects fell out of that, both found by Greptile in review:
- A saved draft can name an adapter the row does not show — one the registry
dropped, or one in the advanced list. The gate asked `sourcePicked`, which
only means "a draft named something", so Next stayed live on a step that
had visibly asked nothing and would hire against the hidden name. It now
asks `sourceSelected`, read off the visible row, and the snap that replaces
an unofferable adapter clears the pick rather than presenting its
replacement as chosen.
- Cmd+Enter bypassed that gate entirely, because the condition was written
out twice and drifted. Both paths now ask one predicate.
Also: selection is a fill rather than a border; the input canvas opens only
for a chosen source; the close control is gone from every step (nothing
downstream of the connect step works until a model is connected); the arc
draws to the design's 424px column with a filled 44px name field; the CTA
reads "Next" through the arc and "Get started" at the end; the login spinner
reads "Preparing..."; the OAuth panels number their fields and show the URL
above the code.
Storybook: the arc stories waited for a button named "Connect" and had been
silently rendering the wrong step since the CTA was renamed. They now wait on
the destination heading. Three e2e specs had the same fault and now select a
source before advancing.
One test was removed rather than replaced — `hydrates again when the same
company comes back through onboarding` drove the close control, and three
substitutes each passed against a wizard with the behaviour deleted. The
reason is recorded where it stood.
Known and not fixed here: SidebarCompanyMenu opens the wizard at step 1 for
"create a new organization", and with no exit an existing user who changes
their mind is trapped.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
## Thinking Path
> - Paperclip manages agents that use different model providers and
adapters.
> - Paperclip must keep agent execution rules clear and predictable.
> - The cheap-model profile added a second execution mode across
adapters, task recovery, APIs, and the UI.
> - That mode increased configuration and recovery complexity.
> - This pull request removes the cheap-model profile as a product
feature.
> - The benefit is one model-selection path for normal work and recovery
work.
## Linked Issues or Issue Description
**What existing behavior does this improve?**
This change simplifies model selection across agent configuration, task
execution, recovery, and adapter capabilities.
**Current behavior**
Paperclip exposes cheap-model profiles in adapter metadata, agent
runtime configuration, task overrides, recovery rules, APIs, and the
board UI. Recovery work can select a different model profile from the
agent's configured model.
**Proposed behavior**
Paperclip uses the agent's configured model for normal work and recovery
work. Status-only recovery stays limited to coordination work. The API
rejects legacy model-profile configuration. A migration removes stored
model-profile values from existing agent, issue, and historical revision
records.
**Reason and benefit**
One model path reduces configuration, API, UI, and recovery complexity.
It also prevents status recovery from becoming a separate product-level
model-routing feature.
**Breaking changes**
This change removes model-profile fields and adapter capability
metadata. Existing stored model-profile values are removed by an
idempotent migration. The validators reject new legacy profile values
with clear errors.
## What Changed
- Removed model-profile types, adapter capabilities, API fields, and
model selection logic.
- Removed cheap-model controls from agent and task UI surfaces.
- Kept status-only recovery limited to coordination context while normal
continuations use the configured agent model.
- Added an idempotent migration that removes stored model-profile values
from agents, issues, and configuration revisions without changing issue
update timestamps.
- Updated tests and product documentation for the single-model behavior.
## Verification
- `pnpm check:token-gates` passes.
- `pnpm -r typecheck` passes.
- `pnpm build` passes.
- `pnpm test:run` completed with 5,607 passing tests and 8
environment-sensitive failures in unrelated fixed-port and
database-deadlock suites. The same failures repeated in an isolated
rerun. CI is the final clean-room result.
## Risks
- This is an intentional breaking change for clients that send
model-profile fields.
- The migration changes legacy agent, issue, and configuration-revision
JSON. It is idempotent and preserves unrelated fields and issue update
timestamps.
- The change is cross-cutting because the removed feature existed in
adapters, shared contracts, the server, plugins, and the UI.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex with `gpt-5`. Reasoning and tool use were enabled. The
runtime did not expose the context-window size.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [ ] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [ ] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - New customers meet it through onboarding, whose last three steps run
inside the tenant: create an agent, connect a model, review
> - The connect step is the one that decides whether the agent can run
at all, and it had drifted — three contributors changed it in parallel,
and its visual language no longer matched the rest of the flow
> - It also could not be looked at without a provisioned stack, so
defects in it were only found by walking a real signup, and the review
step behind it could not be reached at all when it failed
> - This pull request brings the visual work onto the sign-in behaviour
that already shipped, and adds Storybook coverage for all three steps
> - The benefit is that the step is easier to read, and that it can now
be inspected and driven before it ships rather than after
## Linked Issues or Issue Description
**What existing behavior does this improve?**
The onboarding connect-a-model step. It presented the model choice as a
dropdown plus an "Advanced settings" disclosure, and put each credential
type in a different place, so the controls below moved whenever the
choice changed.
**Subsystem affected**
Tenant onboarding wizard (`ui/src/components/OnboardingWizard.tsx`) and
its Storybook coverage.
**Current behavior**
The step offered every registered adapter through a disclosure.
Credential entry appeared in a different shape per source. None of the
three agent-arc steps could be rendered outside a provisioned cloud
stack, so the sign-in panel and the review step were only reachable by
walking a real signup.
**Proposed behavior**
Two brand tiles for the recommended sources, a link that switches
between subscription and API-key credentials, and one canvas that holds
whichever input the current choice needs. Storybook stories mount the
real wizard against fixtures and walk it forward, so every step and its
states can be inspected locally.
**Reason and benefit**
The step reads as one decision rather than three scattered ones, and its
furniture stays still while the choice changes. The stories mean a
regression in it is visible before release instead of during a signup.
**Breaking changes**
No API or schema change. One behavioural narrowing, described under
Risks.
## What Changed
- Replaces the adapter dropdown and "Advanced settings" disclosure with
`ModelSourceTiles` — brand tiles for Claude Code and Codex.
- Adds `CredentialModeLink`, a text toggle between subscription sign-in
and API keys, replacing the disclosure.
- Adds `ConnectInputCanvas`: one surface that holds the sign-in panel or
the API-key field and resizes between them, so the Connect button below
does not move.
- Keeps the existing sign-in behaviour unchanged. `AgentConfigForm`
changes are presentation only — the provider name in the title, the CTA
wording, and `space-y` to `gap`. No change to the login mutations,
queries, or session handling.
- Restores the sleep marks on the dormant agent for the two steps before
the hire.
- Adds Storybook stories for all three agent-arc steps, with fixtures
for the environments, auth signal, both adapters' login flows, and the
hire.
- Copy: names the provider being signed in to ("Sign in to
Anthropic"/"Sign in to OpenAI"), and drops "Clippy" from the agent-name
helper text.
## Verification
- `pnpm vitest run src/components storybook` in `ui/` — 139 tests over
the touched suites, 1986 across `src/components`.
- `pnpm typecheck` in `ui/` — clean.
- Storybook, `Onboarding/Agent arc`: walk each story. Step 1 has no Back
button, steps 2 and 3 do.
- The sign-in gate: on `Connect a model`, press Connect without signing
in. It holds on step 2 and reports "No working authentication was
found." On `Review`, which fixtures an authenticated signal, Connect
reaches the review step.
- Both providers' login flows: press Sign in on the Claude tile for the
authorization URL and browser-code field, and on the Codex tile for the
device URL and code.
- The Claude sign-in was also walked end to end on a staging tenant,
including the OAuth redirect and pasting the code back.
## Risks
- **Onboarding now offers two model sources instead of every registered
adapter.** `ModelSourceTiles` is fed the `recommended` set, which is
`claude_local` and `codex_local`; Gemini, Cursor, Grok, Kimi, OpenCode
and Paperclip Runner are no longer selectable *during onboarding*. This
is deliberate. The full list is unchanged in agent settings, which is
where an adapter can still be switched after the agent exists, and
adding a source back is one `recommended: true` in
`adapter-display-registry.ts`. Flagging it because it is the one
behavioural narrowing here and it is not visible from the diffstat.
- The API key entered on this step is held in component state and
deliberately never written to the onboarding draft, because that draft
is `localStorage`. A customer who leaves mid-step re-enters the key;
that is the intended trade.
- Storybook-only risk: the fixtures now answer the environment test from
the story's auth state. If a future change moves the hire's gate off the
`adapter_auth_missing` check code, the stories would keep passing while
the product regressed. The gate is asserted in the adapter packages' own
tests, not here.
- Motion changes are low risk and reversible: the input canvas animates
its contents only, and its container was deliberately left unanimated
after an animated wrapper clipped the sign-in panel.
## Model Used
Claude Opus 5 (`claude-opus-5`), via Claude Code with extended thinking,
tool use, and browser-driven verification of the Storybook stories.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [ ] I will address all Greptile and reviewer comments before
requesting merge
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent adapters control how Paperclip starts and resumes an agent
runtime.
> - Paperclip Runner is an experimental Rust runtime and must stay
opt-in.
> - The server already rejected new runner selections when the flag was
off.
> - Some setup and onboarding views did not enforce the same boundary.
> - This pull request exposes the existing flag and applies it to every
new setup path.
> - The benefit is a safe rollout with unchanged legacy onboarding and
recoverable existing native runs.
## Linked Issues or Issue Description
**What existing behavior does this improve?**
This improves experimental adapter selection in Settings, onboarding,
new-agent setup, invite setup, and company import.
**Subsystem affected**
Cross-cutting: the React UI and the server onboarding seed service.
**Current behavior**
The server defaulted Paperclip Runner to off, but Settings did not
expose the flag. First-run onboarding could show the runner after
opt-in. A direct new-agent URL and some setup pickers could also reveal
native runner configuration before the availability check completed.
**Proposed behavior**
Settings has a default-off Paperclip Runner toggle. Explicit agent
configuration shows the runner only after the server reports that the
flag is enabled. First-run and invite onboarding always use legacy
adapters. Existing native agents and runs remain readable and
recoverable.
**Reason and benefit**
This keeps the experimental runtime out of normal onboarding. It also
gives administrators one clear opt-in before users can create a native
runner agent.
**Breaking changes**
None. Legacy adapter selection and execution stay unchanged. Existing
native records remain available.
## What Changed
- Added the Paperclip Runner opt-in to Experimental Settings.
- Refreshed adapter availability after the setting changes.
- Kept UI and server-seeded onboarding on legacy adapters.
- Made native runner choices fail closed in new-agent, invite, and
import setup.
- Preserved edit and recovery behavior for existing native agents and
runs.
- Added focused regression tests for flag-off and flag-on behavior.
## Verification
- GitHub Actions will run the repository test, typecheck, build, and
policy gates.
- Focused tests cover Settings, onboarding, agent creation, invite
setup, import setup, and server-seeded onboarding.
- No local test suite was run, per the maintainer request to use GitHub
Actions for verification.
- `git diff --check` passes.
## Risks
Low risk. The change narrows new adapter selection only. The server
remains the final enforcement point. Existing native records do not
depend on the current flag value for read or recovery behavior.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex, GPT-5, with reasoning, tool use, and code execution.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [ ] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Administrators need bounded controls for experimental native
execution.
> - The lower stack adds remote Codex execution and the task workspace.
> - Operators need to configure Codex safely and inspect provider
traces.
> - Unsupported providers must not appear as runnable choices.
> - This pull request adds Codex-only administration and observability.
> - The benefit is a default-off operational surface for production
diagnosis.
## Linked Issues or Issue Description
Refs #12640.
Refs #12616.
Refs #12352.
**Subsystem affected**
Agent configuration, instance experimental settings, run ledger,
provider trace inspector, and administrator actions.
**Problem or motivation**
The native runner lacks one safe operator surface for Codex permissions,
lifecycle, raw trace capture, and run inspection. The integration branch
also contains provider choices that the production backend cannot
execute yet.
**Proposed solution**
Expose only the qualified Codex controls. Keep Paperclip Developer Mode
and runner preview ingress off by default. Gate raw trace actions by
administrator access and existing trace authorization.
**Alternatives considered**
Exposing unfinished providers would create configurations that fail at
runtime. Always-on tracing would increase sensitive data and storage
risk.
**Roadmap alignment**
This work supports governed Cloud and Sandbox agents and production
diagnostics.
## Stack
- Base PR: #12640.
- Lower PRs: #12639 and #12638.
- This PR contains only its 54-file administration and observability
delta.
- This is the final feature PR in the Codex production stack.
## What Changed
- Added Codex-only Paperclip Runner permission and lifecycle controls.
- Added bounded warm idle configuration.
- Kept the provider field fixed to Codex.
- Added administrator-only one-run raw trace requests.
- Added a persistent future-run raw trace toggle.
- Added trace status, metadata, ledger, and canonical runner inspection.
- Added JSON-RPC request-origin grouping and finalization lineage.
- Restored the stateful PRP transcript parser and focused projection
tests required by trace inspection.
- Added default-off Paperclip Developer Mode.
- Added Honeycomb run links for authorized developer mode.
- Disabled the legacy operational skill for `paperclip_runner`.
- Did not expose OpenCode, ACPX, Pi, Claude Managed, or AWS runner
choices.
- Did not change migrations, workflows, dependencies, or
`pnpm-lock.yaml`.
## Verification
- GitHub Actions will run UI tests, server tests, repository typecheck,
build, browser tests, security, and policy gates.
- Tests cover Codex configuration defaults and bounds, administrator
trace actions, persistent settings, ledger inspection, trace lineage,
and Honeycomb links.
- Existing server trace authorization and retention tests remain the
backend authority.
- Local tests were not run. The requested verification policy uses
GitHub Actions for this series.
- `git diff --check runner/task-workspace-experience...HEAD` passes.
- The delta contains 54 files.
## Risks
- Raw provider traces can contain sensitive provider data.
- Existing server authorization controls access, reveal, download,
retention, and deletion.
- The UI gates trace actions by administrator access and developer mode.
- All new instance settings remain off by default.
- Fresh Paperclip Runner configuration remains Codex-only.
- Direct adapters and legacy task behavior do not change in this PR.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex with GPT-5.6. The work used high-reasoning agent mode,
repository tools, GitHub tools, and parallel code-audit agents.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with Fixes: / Closes /
Refs OR (b) described the issue in-PR following the relevant issue
template
- [x] I have not referenced internal or instance-local Paperclip issues
or links
- [x] My branch name describes the change and contains no internal
Paperclip ticket id
- [ ] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - An instance can turn on the `enableManagedSandboxOnly` feature,
which hides the local environment and runs every agent in the
platform-managed environment
> - That feature already gated the environment pickers, the onboarding
wizard, and the server-side run selection, but many other screens still
showed absolute paths on the execution host and still let the user pick
an execution engine
> - On such an instance those controls name a filesystem the user cannot
reach; a path written there is stored and then ignored, which reads as a
broken control
> - This pull request hides the remaining host-path and execution-engine
surfaces behind the same feature, adds a server rule that refuses a
project-workspace path write while the feature is on, and closes a
related route gap in the isolated-workspace pages
> - The benefit is that a managed instance shows no host path and no
folder picker anywhere, and a write that carries a path now fails with a
clear message instead of being silently discarded
## Linked Issues or Issue Description
No public issue exists. The description below follows
`.github/ISSUE_TEMPLATE/enhancement.yml`.
**What existing behavior does this improve?**
The `enableManagedSandboxOnly` instance feature, and the UI surfaces
that show
a host filesystem path: project properties, the new-project dialog, the
project
workspace and execution workspace detail pages, the workspace and task
cards,
plugin local folders, and the agent configuration form with its
per-adapter
fields. It also improves route gating for `enableIsolatedWorkspaces`.
**Subsystem affected**
Cross-cutting (`ui/` and `server/`).
**Current behavior**
When `enableManagedSandboxOnly` is on, the local environment disappears
from the
environment pickers and the server refuses to run an agent on the local
host.
Everything else stays visible. A user still sees:
- the project "Local folder" row, its absolute path, and the
Set/Change/Clear buttons
- the "Local folder" field and its "Choose" folder picker in the
new-project dialog
- the "Local path" field and fact row on a project workspace
- the "Paths" and "Lifecycle commands" groups on an execution workspace
- the working directory on workspace cards, task properties, and runtime
service rows
- the plugin "Local folders" section
- "Working directory (deprecated)", "Command", "Execution engine",
"ACP server command", "ACP state directory", and "Agent instructions
file" in
the agent configuration form
A path typed into any of these names a filesystem no agent on the
instance uses.
The project workspace API also accepts a `cwd` write and stores it.
Separately, `/workspaces`, `/execution-workspaces/*`, and
`/projects/:projectId/workspaces/:workspaceId` render for anyone who
types or
bookmarks the URL, even with `enableIsolatedWorkspaces` off. Only the
sidebar
entry reads that flag.
**Proposed behavior**
With `enableManagedSandboxOnly` on, none of those surfaces render. A
project
whose codebase came from a managed checkout keeps its one-line
"Paperclip-managed folder." label and shows no path. The non-path
controls stay:
repo URL, branch, service URL, port, command output, ACP session mode,
ACP
non-interactive permissions, Codex fast mode, and the sandbox toggles.
The project-workspace create and patch routes, and the nested workspace
on
project create, answer `422` with
"This instance runs agents only in the platform-managed environment;
local
folders are not configurable." when the payload carries a non-null
`cwd`.
A `cwd: null` write still passes, so an instance that just turned the
feature on
can clear a stale path.
With `enableIsolatedWorkspaces` off, the three workspace route groups
redirect to
the dashboard.
**Reason and benefit**
A control that cannot do anything is worse than a missing control: the
user fills
it in, saves, and gets no error and no effect. The server rule turns
that silent
no-op into a clear refusal. The route gate stops a feature that an
instance has
turned off from staying reachable by URL, which is the same standard the
Cases,
Pipelines, and hidden-settings pages already meet.
**Breaking changes**
None for a default instance: both flags are off by default for
self-hosted and
managed instances, so nothing changes unless an operator turns them on.
Stored
`adapterConfig` values are never cleared, so turning the feature off
restores
every previous value.
## What Changed
- Add `ui/src/hooks/useManagedSandboxOnly.ts`, modelled on
`useAppsEnabled`, for
components that do not already read the experimental settings. It
exposes
`hideHostPaths`, which fails closed while the settings query is in
flight, so
a cold cache never flashes a host path before the policy resolves.
Components
that keep their own settings read compute the same gate from
`isFetched`.
- Add `managedSandboxOnly` to `AdapterConfigFieldsProps` and populate it
where
`AgentConfigForm` builds the adapter field props. Resolve the effective
instructions-file gate once as `hideInstructionsFile || hideHostPaths`,
so
every adapter hides that path field with no per-adapter edit.
- Hide under the flag: the project "Local folder" block and its
absolute-path
edit panel (a managed checkout keeps its label, without the path); the
new-project "Local folder" field; the project-workspace "Local path"
field and
fact row; the execution-workspace "Paths" and "Lifecycle commands"
groups; the
working directory on the workspace summary card, the task workspace
card, the
task properties "Folder" row, and the runtime service rows; the plugin
"Local
folders" section; "Working directory (deprecated)" and "Command" in the
agent
form; and the per-adapter "Execution engine", "ACP server command", and
"ACP state directory" for `claude_local`, `codex_local`, and
`gemini_local`.
- Drop two working-directory fallbacks that had no gate to read: the
close-workspace
dialog now falls back to "No additional details", and the reuse-existing
workspace label and picker subtitle fall back to a neutral phrase.
- Refuse a non-null `cwd` with `422` on `POST /projects/:id/workspaces`,
`PATCH /projects/:id/workspaces/:workspaceId`, and the nested workspace
on
`POST /companies/:companyId/projects`, following the
`assertNoAgentHostWorkspaceCommandMutation` precedent on those routes.
- Add `IsolatedWorkspacesRouteGate` and wrap the `/workspaces`,
`/execution-workspaces/*`, and
`/projects/:projectId/workspaces/:workspaceId`
routes with it.
- Leave the SSH "Remote workspace path" and the workspace file browser
alone,
with a comment explaining why.
## Verification
Automated:
- `pnpm --filter @paperclipai/ui exec vitest run` — 479 of 480 files
pass
(4455 of 4456 tests). The one failure is `OnboardingWizard.test.tsx >
renders
instead of throwing when the browser denies storage access`, which also
fails
on `origin/master` and is unrelated to this change.
- `pnpm --filter @paperclipai/server exec vitest run project workspace
instance-settings`
— 45 of 50 files pass. Four files fail on macOS for reasons unrelated to
this
change: `workspace-instance-cleanup`, `workspace-runtime`,
`execution-workspace-runtime-control-conflict`, and
`workspace-runtime-exposure` compare `/var/...` against the resolved
`/private/var/...` or bind real ports. The same files fail on a clean
`master`
checkout on the same machine.
- `pnpm --filter @paperclipai/ui typecheck`
- `tsc --noEmit` in `server/` (after
`pnpm --filter @paperclipai/plugin-sdk ensure-build-deps`). The package
`typecheck` script also builds the Rust runner, which needs `cargo`; it
is not
installed on the machine that ran this.
New and extended tests:
- `ui/src/adapters/managed-sandbox-only-config-fields.test.tsx` — the
three
adapters drop the execution engine, the ACP paths, the instructions-file
path,
and every "Choose" button when the flag is on, and keep the non-path
controls.
- `ui/src/components/AgentConfigForm.render.test.tsx` — flag-on and
flag-off
renders for the working directory, the command, the engine, the ACP
paths, and
the resolved adapter field props.
- `ui/src/components/ProjectProperties.managed-sandbox.test.tsx`,
`ui/src/components/NewProjectDialog.managed-sandbox.test.tsx`,
`ui/src/pages/ProjectWorkspaceDetail.test.tsx`,
`ui/src/components/ProjectWorkspaceSummaryCard.test.tsx`,
`ui/src/components/WorkspaceRuntimeControls.test.tsx`.
- `ui/src/components/IsolatedWorkspacesRouteGate.test.tsx` — redirect
when off,
render when on, and render nothing while the flag query is in flight.
- "Still loading" cases for the project properties, the new-project
dialog, the
workspace summary card, the runtime service rows, and the agent
configuration
form, each asserting that no host path renders before the policy
resolves.
-
`server/src/__tests__/project-workspace-managed-sandbox-routes.test.ts`
— the
`422` on all three write paths, the `cwd: null` pass-through, and the
flag-off pass-through.
Manual check to reproduce: turn on Managed Environment Only in instance
experimental settings, then open a project, the new-project dialog, an
agent's
configuration, and a workspace page. No path, folder icon, or "Choose"
button
appears. Turn the setting off and each control returns with its stored
value.
No documentation change was needed. The operator-facing text for both
settings
lives in the feature catalog entry, which already states the contract
this pull
request now enforces across the UI.
## Risks
- Low. Both flags default to off, so a default instance is unchanged.
- The hidden fields are presentation only. No stored `adapterConfig`
value is
cleared, because an import carries adapter configuration written on
another
instance and clearing it would break that flow. Turning the setting off
shows
every previous value again.
- The `422` is the one behavior change for an API caller, and only while
the
setting is on. `cwd: null` still passes so a stale path can be cleared.
- The route gate renders nothing until the flag query settles, so an
instance
with isolated workspaces on never flashes a redirect. An instance with
the
feature off now redirects a bookmarked workspace URL to the dashboard.
- Every host-path guard fails closed while the settings query is in
flight, so a
default instance shows those controls a moment later than before on a
cold
load. That is the safe direction: the alternative flashes a path a
managed
instance must never show.
- Two path surfaces stay on purpose, each with a comment: the SSH
"Remote
workspace path" is a path on the user's own remote host, and the
workspace file
browser shows workspace-relative paths. The instance Adapters page also
keeps
its "Local path" install option, since that page is an instance-admin
surface
the hosting operator can already hide through the hidden-settings
mechanism.
## Model Used
Claude (Anthropic), Claude Opus, 1M context window, extended thinking,
agentic
tool use through Claude Code.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - The Paperclip Runner now has protocol, provider, tool, package,
persistence, and hidden server boundaries.
> - The server still cannot select that path for a real agent heartbeat.
> - A new runtime must not change any existing direct adapter.
> - An experimental runtime must fail closed when its rollout flag is
off.
> - This pull request adds one guarded Codex vertical slice through
runnerd.
> - The benefit is a production-built runner path that users cannot
start by default.
## Linked Issues or Issue Description
Refs #11962
Refs #12111
Refs #12169
Refs #12176
**Subsystem affected**
Cross-cutting. The change affects the runner package, server
orchestration, shared settings, and adapter configuration UI.
**Problem or motivation**
The hidden PRP coordinator cannot execute a real heartbeat. The
application also needs an explicit rollout boundary before it can expose
the experimental runner. Existing direct adapters must keep their
current execution and finalization behavior.
**Proposed solution**
Add `paperclip_runner` as a Codex-only adapter behind the default-off
`enableNativeRunner` instance flag. Select the native runtime only for
that adapter. Persist the run binding before runnerd starts. Wait for
the durable PRP result and terminal event. Resume the real Codex
provider thread on later heartbeats. Keep persisted native runs readable
and recoverable after the flag changes.
**Alternatives considered**
The server could route `codex_local` through runnerd. That option would
change an existing adapter and weaken rollback safety. The server could
expose all providers now. That option would add unreviewed provider
behavior. The build could depend on a prebuilt runner binary. That
option would make source builds architecture-dependent and difficult to
verify.
**Roadmap alignment**
This work supports the shipped enforced-outcomes, governed-tool, and
self-healing-run milestones. It does not add a new roadmap surface. It
is the guarded execution step after the merged hidden runner boundaries.
**Additional context**
This is the next replacement for the closed large runner pull request.
Task-thread presentation remains a separate follow-up so this change can
preserve the current direct-adapter UI.
## What Changed
- Add `paperclip_runner` as an explicit Codex-only adapter.
- Add the default-off `enableNativeRunner` instance flag.
- Reject fresh create, hire, import, switch, and execution requests
while the flag is off.
- Allow edits to persisted runner agents while the flag is off.
- Recover an already persisted native run even after the flag is
disabled.
- Keep every built-in direct adapter on its existing runtime path.
- Persist an immutable native run binding and revisioned completion
contract before runnerd starts.
- Execute server to PRP to runnerd to Codex to server through the hidden
coordinator.
- Validate the durable result against the terminal event and exact
completion criteria before finalization.
- Preserve the Codex provider thread ID and use `thread/resume` on the
next heartbeat.
- Strip unsupported Codex configuration fields from the experimental
adapter.
- Build a target-native release runner binary from source and vendor it
into the server distribution.
- Install Rust only in the Docker build stage. Do not add a workflow or
lockfile change.
- Stop the runner process group on completion, cancellation, and forced
shutdown.
## Verification
- Run `pnpm --filter @paperclipai/paperclip-runner check:all`. All 69
TypeScript tests and 58 Rust tests pass. Protocol, conformance, replay,
formatting, and generated-file checks pass.
- Run the 12 focused adapter, settings, runtime-selection, coordinator,
direct-isolation, and real Codex integration test files. All 186 tests
pass.
- The real integration test uses PostgreSQL, HTTP, WebSocket, runnerd,
and a fake Codex app server. It proves one `thread/start` followed by
one `thread/resume`.
- Run `pnpm -r typecheck`.
- Run `pnpm build`.
- Run `pnpm check:token-gates`.
- Build the Docker `build` target from a clean context. Confirm that the
server distribution contains an executable `paperclip-runnerd` built
with Debian Rust 1.85.
- Start the server through the source-mode tsx entry point with the
package `dist` directory absent. Confirm the vendor shim resolves source
exports and the server boots.
- Run `pnpm test:run` twice. On this macOS host, 405 files pass and 1
file skips. Eight untouched workspace and loopback tests fail because
macOS resolves `/tmp` and `/var` through `/private` and because
PID-derived test ports exceed 65535. Linux CI must pass the full suite.
- Confirm that the diff contains 52 files. Confirm that it contains no
`.github` or `pnpm-lock.yaml` change.
## Risks
- The feature flag is off by default. A fresh native start fails with a
stable error while the flag is off.
- A persisted native run remains recoverable after the flag changes.
This prevents rollout changes from corrupting recorded work.
- Only local Codex execution is accepted. Other providers and remote
work modes fail closed.
- Existing direct adapters do not start runnerd, create native rows, use
native status arbitration, or enter native finalization.
- The runner receives its one-use bootstrap ticket through the child
environment. The server does not put the ticket in command arguments or
logs.
- The server validates the company, task, agent, run, runner, session,
completion contract, result, and terminal binding before it accepts
completion.
- The build compiles a target-native Rust binary. Cross-platform release
packaging remains a later concern. Source builds and Docker builds
compile for their current target.
- Docker needs enough build memory for the existing server TypeScript
compile. The Docker build stage sets a 4 GB V8 heap limit.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex with GPT-5. The exact deployment ID and context-window
size are not exposed. The model used agentic reasoning, repository
tools, code execution, and test execution.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and applicable tests pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Agent adapters use provider-specific login flows
> - Codex device login needs a live pseudo-terminal (PTY), while the
shared channel still uses Claude-specific names
> - The old streamed-exec path does not provide the prompt transport
that Codex needs
> - This pull request moves Codex device login to the shared login PTY
and removes the dead streamed-exec path
> - The benefit is one controlled login transport with fail-closed
capability checks and safer credential reads
## Linked Issues or Issue Description
**Problem or motivation**
Codex device login used a streamed-exec path that did not provide the
required prompt transport. The shared login channel also exposed
Claude-specific names outside Claude code.
**Expected behavior**
The host selects a fixed login command from trusted adapter data. Codex
login uses the provider login PTY. Providers without that capability
fail closed.
**Proposed solution**
Use a server-controlled session home, create and validate it as a fresh
0700 directory, read credentials from one validated descriptor, and
rename shared channel names to the neutral login PTY family.
**Alternatives considered**
Keep the shared login PTY as the single transport. Do not keep the
removed streamed-exec path because it cannot provide the required prompt
transport.
**Roadmap alignment**
This change supports the planned login transport work. It does not add a
separate roadmap item.
## What Changed
- Route Codex device login through the shared login PTY transport.
- Select the login command from a closed internal command key.
- Carry a server-controlled session home through the launch contract.
- Create and validate the session home as a fresh 0700 directory owned
by the login user.
- Read the credential file with descriptor-relative, no-follow path
walking and final descriptor checks.
- Gate the login route and run lease on the provider login PTY
capability.
- Rename shared channel names to the neutral login PTY family.
- Remove the streamed-exec transport value, selector field, driver
branch, and related tests.
- Hide Codex login in the user interface when the provider lacks the
login PTY capability.
## Verification
- Server unit suites pass: 89/89.
- Adapter-utils suites pass: 262/262.
- Codex-local suites pass: 326/326.
- Credential-read reader suite passes: 20/20.
- Daytona login PTY suite passes: 30/30.
- Device-login suites pass: 56/56.
- TypeScript checks pass for server, adapter-utils, and UI.
- GitHub Actions must pass after pull request creation.
- Greptile review must reach 5/5 with no open P2 findings,
recommendations, or follow-ups.
## Risks
- Providers without a login PTY capability lose Codex login support by
design.
- The credential read rejects invalid ownership, mode, type, path, and
size.
- The launch-time sandbox directory race remains outside the threat
model because the login runs inside the sandbox and a hostile sandbox
already controls its credential.
## Model Used
OpenAI Codex, GPT-5, tool use and code review assistance. The exact
context window and reasoning mode are not exposed by the runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Managed deployments provision a platform-managed default environment
for agent runs; the UI shows this environment in selectors, the agent
form, run details, and the environments page
> - Those surfaces append the raw driver key to the environment name, so
users see labels like "Paperclip Computer (sandbox)", "Paperclip
Computer · sandbox", and fallback copy such as "Managed sandbox" and
"The sandbox has no ready authentication"
> - "sandbox" is infrastructure vocabulary, not the product name of the
environment; showing it next to the managed environment's name is
confusing and off-brand
> - This pull request renders platform-managed environments by name
alone and rewords the sandbox-phrased copy, while user-created
environments keep the driver suffix so mixed lists stay distinguishable
> - The benefit is that the default environment reads as one clear
product name everywhere, and self-hosted users lose nothing: their own
environments still show the driver
## Linked Issues or Issue Description
**What existing behavior does this improve?**
Display of the platform-managed default environment across the UI.
**Subsystem affected**
UI (environment selectors, agent config form, environments page, agents
page, run details) and the claude-local/codex-local adapter auth checks.
**Current behavior**
The agent form labels the inherited default environment as "Name
(sandbox)". Environment selectors and the environments list render "Name
· sandbox". The agents page describes the environment as "<provider>
sandbox provider". The agent form's fallback label is "Managed sandbox".
Adapter auth checks say "The sandbox has no ready authentication for
this adapter."
**Proposed behavior**
Platform-managed environment rows (`metadata.managedByPaperclip`) render
their name alone. The fallback label is "Paperclip Computer". The agents
page describes managed environments as "Managed by Paperclip". Run
details omit the driver suffix for sandbox-driver environments (the
adjacent Provider entry already identifies the mechanism). Adapter auth
checks say "This environment has no ready authentication for this
adapter."
**Reason and benefit**
The managed environment carries a product name. Appending the raw driver
key ("sandbox") to it is noise and contradicts the product naming.
User-created environments keep the driver suffix, so mixed lists stay
distinguishable.
**Breaking changes**
None. Message text of the auth check is not read programmatically; the
UI keys off `ADAPTER_AUTH_MISSING_CHECK_CODE`. Rows without the managed
marker render exactly as before.
## What Changed
- New `environmentDisplayLabel` helper in
`ui/src/lib/managed-sandbox-environment.ts`: managed rows → name alone;
other rows → "Name · driver".
- `AgentConfigForm`: inherited-default label uses the helper; fallback
copy "Managed sandbox" → "Paperclip Computer"; environment options use
the helper.
- `ProjectProperties`, `CompanyEnvironments`: environment selector
options use the helper; the environments-list row hides the driver
suffix on managed rows; the managed detail page's fallback description
no longer says "sandbox".
- `Agents` page: managed environments are described as "Managed by
Paperclip" instead of "<provider> sandbox provider".
- `CommentThread` run details: the driver suffix is omitted for
sandbox-driver environments.
- claude-local and codex-local adapters: auth-missing check message/hint
reworded from "sandbox" to "environment" (ACP and environment-test
paths); claude-local probe/effort/login hints reworded the same way.
- Run status lines: "Syncing workspace to sandbox", "Exporting git
changes from sandbox", "Starting adapter in sandbox", and friends now
say "environment"; "Finalizing sandbox workspace" → "Finalizing
workspace". Templated transfer-progress lines map the `sandbox`
transport key to "environment" for display (`runtime-progress.ts`).
- Agent form sign-in panel: "Sign in to the sandbox" → "Sign in to the
environment"; "Authenticated. The sandbox has credentials now." → "…The
environment has credentials now."
- Feature catalog + instance settings card: "Managed Sandbox Only" →
"Managed Environment Only" (setting key unchanged; the card keeps its
alphabetical slot).
- Server agents routes: execution-target failure and test-identity copy
no longer say "sandbox"; workspace-mode label "Cloud sandbox" → "Cloud
environment".
- Tests: new `environmentDisplayLabel` unit cases; new `AgentConfigForm`
render case asserting the managed default renders without "(sandbox)" or
"· sandbox"; status-line assertions updated across adapter-utils, server
heartbeat/live-run, and UI chat suites.
## Verification
- `pnpm --filter @paperclipai/ui typecheck` — clean.
- `pnpm --filter @paperclipai/adapter-claude-local typecheck` and
`--filter @paperclipai/adapter-codex-local typecheck` — clean.
- `vitest run` for `managed-sandbox-environment.test.ts`,
`AgentConfigForm.render.test.tsx`, `CompanyEnvironments.test.tsx`,
`Agents.test.tsx`, `CommentThread.test.tsx`, `NewAgent.test.tsx` — all
green (118 tests across the two runs).
## Risks
Low risk. Cosmetic label changes only; no data or API changes. Rows
without `metadata.managedByPaperclip` render exactly as before, so
self-hosted deployments with their own environments see no change. The
only self-hosted-visible wording changes are the adapter auth-check
message and the driver suffix omission on sandbox-driver rows in run
details.
## Model Used
- Claude (Anthropic) — claude-fable-5 (Claude Fable 5), Claude Code CLI,
extended thinking, tool use.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes (no
docs reference these labels)
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Adapter Test checks whether an agent adapter can run with its
configured environment, and every local-driver adapter (Claude, Codex,
Gemini, OpenCode, Pi, Cursor, etc.) shares this Test route and its UI
resolution logic
> - The Claude ACP Test lane could report pass without checking local or
remote authentication, and the shared Test route and UI had gaps in
environment binding, probe safety, and managed-sandbox resolution that
affect every adapter that uses the Test button, not only Claude
> - This pull request verifies authentication on every Claude ACP
target, and closes the shared Test-route/UI gaps: tenant-binding on the
route, a managed-sandbox-only redirect that matches the real run path,
and a three-tier environment resolution in the UI
> - The benefit is a truthful Test result with safer probe execution and
tenant isolation, for Claude specifically and for every other local
adapter that shares this Test surface
## Linked Issues or Issue Description
**What happened?**
The Claude ACP Test lane returned `status: "pass"` without checking
authentication for some local and non-sandbox targets. Separately, the
shared `/companies/:companyId/adapters/:type/test-environment` route —
used by every local-driver adapter, not only Claude — accepted a foreign
environment id, and its UI resolution did not mirror the server's
managed-sandbox-only redirect.
**Expected behavior**
The Test lane checks the resolved credential and hello probe for every
Claude ACP target. The shared adapter Test route rejects a foreign
environment before it reveals environment details or starts a lease, for
any adapter type. The Test's environment resolution (UI and server)
matches the real run's three-tier resolution, including the
managed-sandbox-only redirect.
**Steps to reproduce**
1. Run the Claude ACP Test lane against a local target without a valid
credential.
2. Run the adapter Test route with an environment id from another
company (any adapter type).
3. Observe the pass result on step 1, or the missing tenant-binding
rejection on step 2.
**Paperclip version or commit**
`933749e01f74e82ce5d315c071be534d04e01158`
**Deployment mode**
Local dev (`pnpm dev`) and server route tests.
**Agent adapter(s) involved**
Claude Code directly (the ACP auth-verification work). The
tenant-binding guard, managed-sandbox-only redirect, and UI three-tier
resolution apply to the shared adapter Test route and affect every
local-driver adapter (Codex, Gemini, OpenCode, Pi, Cursor, etc.), not
only Claude — see "What Changed" below for the split between Claude-only
and shared changes.
**Database mode**
Not database-related.
**Access context**
Both board and agent paths use the affected Test surface, for every
local-driver adapter.
**Additional context**
Two commits that were previously bundled into this PR — a
`plugin-worker-manager` duplex-channel frame-bound fix and a
`workspace-runtime` exit-persist crash fix — are unrelated to the
adapter Test lane and have been split out into their own PRs: #11860 and
#11861.
## What Changed
Claude-only (`packages/adapters/claude-local`):
- Verify `CLAUDE_CODE_OAUTH_TOKEN` and run the hello probe for every
Claude ACP target.
- Keep `adapter_auth_missing` sandbox-only and report missing
non-sandbox credentials as a warning.
- Add a deny-by-default probe environment builder for the ACP and CLI
local probes.
- Log only fixed probe context and allowlisted classifications.
- Seed the host OAuth token into the hello probe environment.
Shared, cross-adapter (`server/src/routes/agents.ts`,
`ui/src/lib/adapter-test-environment.ts`,
`ui/src/components/AgentConfigForm.tsx`,
`ui/src/components/OnboardingWizard.tsx`):
- Add a company-binding guard and a binding assertion for the generic
`/companies/:companyId/adapters/:type/test-environment` route, so a
foreign-company environment id is rejected before any secret resolution
or sandbox lease, for every adapter type.
- Resolve all three server environment tiers (agent default, instance
default, local default) in the UI, and add the managed-sandbox-only
redirect so the Test probes the same target a real run would use.
- Enforce onboarding Test results: block hire on a failed environment
test.
- Add regression tests for authentication, tenant binding, probe safety,
diagnostics, and UI resolution.
## Verification
- Adapter suites pass for the Claude local server probe, remote, ACP,
auth, probe environment, and config paths.
- Server route tests pass, including the five tenant-binding cases.
- UI adapter Test environment resolver tests pass for all three
resolution tiers.
- Adapter package `tsc --noEmit` exits 0.
- Full CI must pass on this pull request.
## Risks
The probe environment now denies caller variables by default. A required
variable that is not on the allowlist could stop a probe from starting.
The route now rejects foreign environment ids with a fixed 403 response.
The managed-sandbox-only redirect changes where the Test (and the login
affordance) probes for every local-driver adapter under that policy, not
only Claude — operators running other local adapters under
managed-sandbox-only will see their Test target move from local to the
managed sandbox, matching what real runs already do. The change limits
secret and diagnostic exposure.
## Model Used
Original implementation: OpenAI Codex, GPT-5; exact context window not
exposed in that run; tool use and code execution.
This revision (commit split and title/description correction): Claude,
Sonnet 5 (claude-sonnet-5). The original title and description described
this PR as Claude-only; review found it also changes the shared adapter
Test route and UI resolution used by every local-driver adapter, and
carried two unrelated server fixes. Claude split those two commits into
#11860 and #11861 via `git rebase --onto` (verified byte-identical to
the original tree minus those commits) and rewrote this description to
reflect the actual scope. No functional code in this PR was authored by
Claude.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Local agent adapters (`claude_local`, `gemini_local`, `grok_local`,
…) are the integration surface that lets Paperclip run coding CLIs on
the host machine
> - The Kimi Code CLI (`kimi`, Moonshot AI) has a documented
non-interactive mode, `kimi -p --output-format stream-json` with session
resume via `kimi -r`, but Paperclip has no built-in adapter for it
> - So Kimi users (especially Kimi membership / OAuth subscribers)
cannot onboard their CLI to Paperclip agent teams
> - This pull request adds a complete built-in `kimi_local` adapter
(both execution engines, session management, instructions + skills
delivery, thinking-effort control, environment test, UI and CLI modules,
docs) following the established `gemini_local`/`grok_local` package
pattern
> - Kimi Code ships an ACP server (`kimi acp`), so the adapter runs on
Paperclip's shared acpx engine by default (streaming transcript with
live tool status, like `claude_local`/`gemini_local`) and falls back to
a headless CLI lane (`kimi -p --output-format stream-json`) when ACP
prerequisites are unavailable
> - The benefit is that Kimi Code becomes a first-class Paperclip agent
lane: selectable in the UI, resumable across heartbeats, with the same
operating context (instruction bundle, skills, effort) and streaming
transcript the other local adapters get
## Linked Issues or Issue Description
- Supersedes #9880 (same branch; expanded from the CLI-only lane into a
complete adapter with the default ACP engine lane, control-plane skill
install, and live transcript wiring)
- Refs #9879 (adapter request for Kimi Code CLI, filed with this PR)
- Refs #163 (original Kimi support request)
Duplicate/related prior PRs, per the dedup search (both appear stale: no
updates or maintainer review since May 2026, and both target an older
Kimi CLI interface; calling them out for reviewer context per
CONTRIBUTING.md):
- Refs #6276 (`feat: add kimi-local adapter`): targets an older
array-based content format (`{type: think}`/`{type: text}` blocks), not
the current documented stream-json schema
- Refs #5202 (`feat(adapter): add Kimi CLI local adapter with Wire
protocol support`): builds on a `--wire` JSON-RPC interface that current
Kimi Code CLI (0.27.0) no longer documents; the current documented
headless interface is `-p --output-format stream-json`
This PR is a fresh implementation against current master and the
currently documented/verified Kimi CLI behavior (see Verification).
Happy to fold in anything useful from the earlier attempts if a reviewer
prefers.
## What Changed
- **New adapter package** `packages/adapters/kimi-local`
(`@paperclipai/adapter-kimi-local`), modeled on
`gemini-local`/`grok-local`:
- `src/server/execute.ts`: spawns `kimi -p <prompt> --output-format
stream-json` (argv array, no shell), `-m <model>` only when configured,
`-r <sessionId>` when the stored session cwd matches the run cwd,
automatic fresh-session retry on unrecoverable-session errors,
headless-safe env (`CI=1`, `NO_COLOR=1`, `KIMI_CODE_NO_AUTO_UPDATE=1`,
`TERM=dumb`; user-configured values win), full remote (ssh/sandbox)
execution lane with runtime install via `@moonshot-ai/kimi-code`
- **Instruction bundle delivery**: the prompt path directive now names
the sibling instruction files (`./HEARTBEAT.md`, `./SOUL.md`,
`./TOOLS.md`) alongside the prepended entry file, and local runs pass
`--add-dir <instructions-dir>` so Kimi can actually open them (matching
`claude_local`). Without this, only the entry file reached Kimi and
agents improvised the operating workflow that `HEARTBEAT.md` documents
- **Thinking effort**: a configured `effort` is forwarded as the
`KIMI_MODEL_THINKING_EFFORT` operational override (Kimi has no
per-invocation effort flag). It is only sent for models that advertise
`support_efforts` (currently `kimi-code/k3`) to avoid provider
rejections, and `medium` maps to `high` since Kimi has no medium tier
(`low`/`high`/`max` pass through)
- **Skills delivery**: desired Paperclip skills are delivered via Kimi's
`--skills-dir` flag from a dedicated per-run directory (a local
snapshot, or the synced snapshot on remote targets), so skills load
reliably and in isolation. Paperclip never overwrites the shared
`$KIMI_CODE_HOME/skills` home, so skills installed by the operator or
other agents are left intact. `--skills-dir` is only passed when at
least one skill is desired, so unconfigured agents keep Kimi's default
skill discovery
- **Live run status**: the adapter now forwards each streamed
stream-json line to `onEvent` (assistant `content` as an assistant
snippet, `tool_calls` as tool-name events), which drives the
issue-thread activity indicator (`currentToolName` /
`lastAssistantSnippet` / `lastEventAt`). Previously the adapter only
wrote the raw run log, so the issue thread showed a stale "no output for
N s" line with no tool or reasoning context while Kimi worked. Tool
results are omitted so the last meaningful "Using X" / snippet is not
overwritten by a generic label
- `src/server/parse.ts`: parses the verified Kimi stream-json event
shapes (`assistant` text, `assistant.tool_calls` with JSON-string
arguments, `tool` results, trailing `meta.session.resume_hint` for
session-id capture) plus failure classifiers (`kimi_auth_required`,
transient network, unrecoverable session). A signaled exit (null exit
code, not a timeout) is now reported as a failure rather than coalesced
to success, and the error message names the terminating signal
- `src/server/skills.ts`: lists/syncs Paperclip skills for the adapter's
skill-management surface
- `src/server/test.ts`: environment test covering CLI resolution + `kimi
--version`, cwd check, auth detection (OAuth credential dirs, keyed
`[providers.*]` in config.toml, or the `KIMI_MODEL_NAME` +
`KIMI_MODEL_API_KEY` env pair), and a live hello probe
- `src/ui/` (stdout-line parser for transcripts, config builder) and
`src/cli/` (stream event formatter) modules
- Root metadata: three managed model aliases
(`kimi-code/kimi-for-coding`, `kimi-code/kimi-for-coding-highspeed`,
`kimi-code/k3`), effort-capable-model metadata (`EFFORT_CAPABLE_MODELS`,
effort mapping helpers), `agentConfigurationDoc`
- Tests: 101 tests across parse, execute (args building, resume gating,
retry, auth error code, timeout, signaled-exit failure, effort
forwarding/gating/mapping, `--add-dir` instructions directive,
`--skills-dir` gating, `onEvent` runtime-event forwarding), ACP engine
(engine resolution, acpx config build, node-version gate), ACP
transcript delegation, environment test, UI parse/build-config
- **ACP engine lane (default)** (`src/server/acp.ts` + shared
`adapter-utils/acpx-engine`): Kimi Code ships an ACP server (`kimi
acp`), so `kimi_local` now runs on Paperclip's shared acpx engine by
default, matching `claude_local`/`codex_local`/`gemini_local`. The
issue-thread transcript streams live (assistant text deltas, tool calls
with a `pending`->`completed` status lifecycle) instead of the CLI
lane's bursty complete-message output. Registered `kimi_local -> "kimi"`
in `ACPX_ADAPTER_AGENT_IDS` and resolved the built-in agent command to
`kimi acp`; `execute.ts` dispatches to the ACP executor first with an
automatic CLI fallback when ACP prerequisites fail (`engine=acp`
requires ACP, `engine=cli` pins the headless lane); `index.ts` falls
back to the shared acpx session codec; the UI/CLI delegate `acpx.*`
events to the shared acpx transcript parser and event formatter. The
headless CLI lane (above) remains as the fallback
- **Registration** (one entry each, mirroring existing adapters): server
adapter registry + `BUILTIN_ADAPTER_TYPES`, `AGENT_ADAPTER_TYPES`
(shared), UI adapter registry + display registry (`Kimi Code`, Moon
icon) + capabilities defaults, CLI adapter registry, `Dockerfile`
(package copy + `npm install --global @moonshot-ai/kimi-code@latest`),
`vitest.config.ts` workspace, `scripts/release-package-manifest.json`
- **Behavioral sets** mirroring `gemini_local` (Kimi resumes sessions
the same way): `GIT_SENSITIVE_LOCAL_ADAPTER_TYPES`,
`SESSIONED_LOCAL_ADAPTERS` (heartbeat + recovery),
`REMOTE_MANAGED_ADAPTERS`, ssh/sandbox execution-target allow-lists,
`ADAPTER_DEFAULT_RULES_BY_TYPE` (`timeoutSec: 0`, `graceSec: 15`), and
`LEGACY_SESSIONED_ADAPTER_TYPES` + `ADAPTER_SESSION_MANAGEMENT` in
adapter-utils
- **UI touch-points**: New Agent default-model branch, AgentConfigForm
command map (`kimi_local: "kimi"`) + model defaults + a Kimi-specific
thinking-effort option list (`Low`/`High`/`Max`, reflecting Kimi's tiers
rather than borrowing Claude's), OnboardingWizard (command map, model
default, `kimi login` / `KIMI_MODEL_NAME + KIMI_MODEL_API_KEY` auth
hints, manual-debug command line), InviteLanding enabled adapters
- **Control-plane skill install** (`cli/src/commands/client/agent.ts`):
`paperclipai agent local-cli` seeded the Paperclip control-plane skills
into `~/.codex/skills` and `~/.claude/skills` so Codex/Claude agents
auto-discover the API reference every run. Kimi had no equivalent
target, so `kimi_local` agents began each session without the
control-plane skill and rediscovered routes (e.g. the company-scoped
`POST /api/companies/{companyId}/issues`) by trial and error. Added
`~/.kimi-code/skills` (honoring `KIMI_CODE_HOME`) as a third install
target for parity. Independent of the per-run `--skills-dir` delivery,
which only applies to explicitly configured skills.
- **Docs**: `docs/adapters/kimi-local.md` (prerequisites, auth options,
config fields including `effort`, session resume, instruction bundle,
skills delivery, control-plane skill install) + a row in
`docs/adapters/overview.md`
Out of scope (deliberately): model profiles, built-in agent
`allowedAdapterTypes` additions.
## Verification\n\nCurrent-master rebase verification (OpenAI Codex,
2026-08-03): 13 focused files / 231 tests pass; adapter-utils, server,
UI, CLI, and Kimi adapter typechecks pass; full repository build and UI
token gates pass. The branch is conflict-free against master at head
`1249df117c5e12e5771b9a570a6340866450619e`.\n\nAutomated (all from repo
root, pnpm 9.15.4, Node 22):
- `vitest run packages/adapters/kimi-local`: 89/89 pass (includes
coverage for the instruction `--add-dir` directive, effort
forwarding/gating/mapping, `--skills-dir` gating, the signaled-exit
failure path, and `onEvent` runtime-event forwarding with cross-chunk
line buffering)
- `vitest run server/src/__tests__/adapter-registry.test.ts
server/src/__tests__/adapter-routes.test.ts
server/src/services/heartbeat-stop-metadata.test.ts
ui/src/adapters/adapter-display-registry.test.ts`: 37/37 pass
- `vitest run cli/src/__tests__/skills.test.ts`: 13/13 pass (the
control-plane skill install target follows the existing Codex/Claude
install path, whose symlink logic is unchanged)
- `vitest run packages/shared`: 307/307 pass; `vitest run
packages/adapter-utils`: pass except one pre-existing, unrelated failure
(`mcp-isolation.integration.test.ts` requires Claude CLI ≥ 2.1.207; host
has 2.1.185, fails identically on unmodified master)
- `pnpm --filter @paperclipai/adapter-kimi-local typecheck|build`, plus
typecheck of `server`, `ui`, `cli`, `adapter-utils`: all clean
- `pnpm install --frozen-lockfile`: passes (the PR diff itself contains
no lockfile changes, per repo policy; verified against a locally
regenerated lockfile)
- `node scripts/check-no-git-push.mjs` and `node
scripts/check-forbidden-tokens.mjs`: pass
- CI note: the `policy` job's release-bootstrap step is expected to stay
red until a maintainer bootstraps the first npm publish of
`@paperclipai/adapter-kimi-local`; see the CI Note for Maintainers
comment. All other contributor-actionable checks are green.
Manual end-to-end (real Kimi CLI 0.27.0, OAuth login, dev server on an
isolated instance):
1. Server `GET /api/adapters` lists `kimi_local` as builtin with correct
capability flags; models endpoint returns the three Kimi models
2. `POST .../adapters/kimi_local/test-environment`: all checks pass,
including a live `kimi -p` hello probe
3. Created a `kimi_local` agent and invoked two heartbeats: run 1
spawned `kimi -p ... --output-format stream-json`, Kimi used its `Read`
tool, produced the expected answer, and the session id was captured from
the `session.resume_hint` meta event; run 2 resumed the **same** Kimi
session (`sessionIdBefore == sessionIdAfter`) via `-r`
4. UI: adapter appears in the New Agent dropdown; selecting it shows the
Kimi command placeholder, the three models, and the Kimi config fields;
the run transcript renders Kimi tool calls via the adapter's stdout
parser
The instruction-bundle, thinking-effort, and `--skills-dir` changes
landed after the manual run above. They are covered by the unit tests
listed under Automated, and the Kimi CLI flags they rely on
(`--add-dir`, `--skills-dir`, `KIMI_MODEL_THINKING_EFFORT`) were
confirmed against the installed Kimi Code CLI 0.27.0 (`kimi --help`,
config-file thinking-effort docs).
Screenshots (assets branch on the fork, not part of the diff):
## Risks
- Low risk to existing behavior: the change is additive, one new
workspace package plus single-entry registrations alongside existing
adapters; no existing adapter code paths are modified.
- The adapter invokes the locally installed `kimi` CLI; like other local
adapters, run behavior depends on the host's Kimi version. The parser is
written against the documented/verified 0.27.0 stream-json schema and
degrades gracefully (malformed lines are skipped, failures surface as
run errors).
- `--skills-dir` overrides Kimi's auto-discovery of user and project
skills for the run. This is intentional (paperclip-managed agents get a
reproducible, isolated skill set), and it is only passed when at least
one Paperclip skill is desired, so unconfigured agents keep default
discovery.
- Thinking effort is only forwarded to models that advertise
`support_efforts` (currently `kimi-code/k3`); `EFFORT_CAPABLE_MODELS`
must be extended when more Kimi models gain support, otherwise a
configured effort is silently ignored for them.
- `Dockerfile` now installs `@moonshot-ai/kimi-code@latest` globally
alongside the other agent CLIs, so image size increases slightly.
- Maintainer action needed for the npm bootstrap gate: the `policy`
job's release-bootstrap step fails until the first npm publish of
`@paperclipai/adapter-kimi-local` (the gate from #5146 that every new
adapter package has passed through). Enrollment with `publishFromCi:
true` is required by the manifest validator (dropping the entry,
`false`, or `private` are all rejected), so this is intentionally left
to a maintainer. Remaining CI lanes are expected to run once it is done.
## Model Used\n\n- **Current-master rebase, conflict adaptation, and
registry-parity coverage:** OpenAI, **GPT-5 Codex** (Codex agent; exact
serving model ID and context-window size were not exposed to the
runtime), with repository, shell, Git, and GitHub tooling. It preserved
Hawik’s commit authorship, reconciled ACPX and environment-capability
changes, added current registry tests, and ran the verification
above.\n- **Adapter implementation and initial review:** Moonshot AI,
**Kimi K3 Coding** (latest), via **Kimi Code CLI v0.27.0**
(`kimi-code/k3` alias, 1M-token context window, thinking mode, agentic
tool use). The CLI agent explored the repo, wrote the adapter
implementation (delegated to a coder sub-agent of the same model), ran
tests, and drafted the first version of this PR body. A second
model-driven review pass (read-only, same model) audited the diff for
security/correctness before submission; its findings (shell-quoting
hardening, auth-detection false positive, session-compaction
registration, test gaps) were fixed and are included.
- **Harness-context fixes and review responses:** Anthropic, **Claude
Opus 4.8** (`claude-opus-4-8`) via Claude Code. Diagnosed from run logs
that Kimi received only the entry instructions file (not the
`HEARTBEAT.md`/`SOUL.md`/`TOOLS.md` bundle) and that `effort` was never
wired, then implemented the instruction `--add-dir` delivery,
`KIMI_MODEL_THINKING_EFFORT` forwarding, and `--skills-dir` skill
delivery, added the accompanying tests and docs, and addressed the
automated review comments (preserving external skills on remote sync,
treating a signaled exit as a failure). Also extended the `paperclipai
agent local-cli` installer to seed the control-plane skills into
`~/.kimi-code/skills` for Codex/Claude parity, wired `onEvent` runtime
events so the issue-thread activity indicator reflects Kimi's tool and
reasoning output live, and built the ACP engine lane (`kimi acp` via the
shared acpx engine, default) so the transcript streams with live tool
status like the other ACP adapters. The Kimi CLI flags, subcommand, and
env var relied on here were verified against the installed Kimi Code CLI
0.27.0.
- All CLI behaviors claimed here (`-p`, `--output-format stream-json`,
`-r` resume, event shapes, `--add-dir`, `--skills-dir`,
`KIMI_MODEL_THINKING_EFFORT`) were verified empirically against the
installed Kimi CLI, not assumed.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green *(only the release-bootstrap step
remains red, pending the maintainer npm publish described in Risks)*
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
*(will address all Greptile comments as they arrive)*
- [x] I will address all Greptile and reviewer comments before
requesting merge
---
## Maintainer Addendum (2026-08-20)
The shared acpx-engine and issue-chat changes (run-summary segmentation,
placeholder tool-event coalescing,
`ISSUE_CHAT_TRANSCRIPT_MAX_VISIBLE_ENTRIES` 30 → 400, live-reasoning UI)
have been **extracted to #11761** so the cross-adapter behavior changes
review and revert independently — both commits there preserve @hawikk's
authorship. This PR is now the kimi-specific adapter only (60 files,
+3,793/−8, essentially pure addition); the only shared-engine touch left
is the `kimi acp` command resolution. `publishFromCi` is `true` — the
package name is bootstrapped on npm.
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Paperclip <noreply@paperclip.ing>
Co-authored-by: Dotta <bippadotta@protonmail.com>
Co-authored-by: Devin Foley <devin@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Sandbox agents need a safe login path for each supported adapter
> - Codex device login and Claude setup-token login used separate
session stores and route logic
> - Separate stores made session lookup, expiry, and login capability
checks harder to keep consistent
> - This pull request unifies both flows on one session table and one
capability contract
> - The benefit is one company-scoped login model with public session
identifiers and shared lifecycle rules
## Linked Issues or Issue Description
**Subsystem affected**
Cross-cutting (multiple of the above)
**Problem or motivation**
Codex and Claude sandbox login used separate session stores and
different route paths. This split increased the risk of inconsistent
company scoping, session lookup, and cleanup.
**Proposed solution**
Use `adapter_auth_sessions` for both login flows. Use public session
identifiers for API access. Select login behavior from projected adapter
capability data. Share the route spine, lease arguments, runner
lifecycle, and reaper rules.
**Alternatives considered**
Keep two session tables and add matching fixes to both routes. This
keeps duplicate logic and does not provide one capability contract, so
this pull request uses shared infrastructure.
**Roadmap alignment**
This change supports the shipped Cloud / Sandbox agents milestone in
`ROADMAP.md`.
## What Changed
- Unify Codex device login and Claude setup-token login on
`adapter_auth_sessions`.
- Return and look up sessions with company-scoped public session
identifiers.
- Enforce one active session for each company, owner, and adapter.
- Share the login route spine, sandbox lease arguments, runner
lifecycle, and missing-auth check.
- Add a standalone setup-token reaper with adapter-specific row
selection.
- Add optional login capability projection for adapters and drive route
and UI selection from that data.
- Rename the provider flag to `supportsLoginPty` and validate its
deprecated alias.
- Remove the old Claude setup-token session table and add the required
migrations.
## Verification
- Server typecheck passed with `tsc`.
- Database typecheck passed.
- UI typecheck passed with `tsc -b`.
- Codex login service and route suites passed.
- Setup-token session, route, and reaper suites passed.
- Adapter session schema, plugin validator, capability projection, UI
render, and Daytona suites passed.
- GitHub Actions must confirm the complete CI gate after pull request
creation.
## Risks
- The migrations remove short-lived in-flight login rows during
deployment. A login that spans the migration can continue until its
provider lease expires.
- The Codex credential store remains company-scoped. A cross-owner
credential race remains a documented, board-accepted risk.
- API clients that use internal session row identifiers no longer work.
The API accepts only public session identifiers.
## Model Used
Codex, GPT-5, exact runtime model ID not exposed in this handoff, large
context window, reasoning, and repository tool use. The implementing
engineer produced the code with AI assistance.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` /
`Closes: #` / `Refs: #` OR (b) described the issue in-PR following the
relevant issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - The Claude local adapter supports subscription login through a
sandbox
> - The new-agent page must show login before the user creates an agent
> - Test results must not expose raw sandbox diagnostics or secret
values
> - This pull request adds the login UI to both Test lanes and closes
the diagnostic boundary
> - The branch also adds durable cleanup recovery for failed sandbox
teardown
> - Reusable sandboxes must retain both their recorded teardown
configuration and a valid lifecycle path until destruction succeeds
> - The benefit is a usable login flow with fixed public checks,
redacted server logs, and recoverable sandbox cleanup
## Linked Issues or Issue Description
Related public work:
[#9488](https://github.com/paperclipai/paperclip/pull/9488) adds
first-class recognition for `CLAUDE_CODE_OAUTH_TOKEN` in headless and
remote runs. Related public issue:
[#2681](https://github.com/paperclipai/paperclip/issues/2681) requests
Claude Code subscription support. This pull request adds the login
transport and new-agent UI flow that those changes do not provide.
**Subsystem affected:** Claude local adapter, server login probes,
sandbox provider setup, cleanup recovery, and the new-agent UI.
**Problem or motivation:** The Test lanes did not show the sandbox login
panel in all supported cases. Test results also exposed raw probe
diagnostics, and JSON escapes could end secret redaction early.
**Proposed solution:** Surface the login capability through the bundled
provider manifest. Prepare the same probe runtime in the ACP lane. Send
diagnostics only to redacted server logs. Keep Test checks on fixed
public messages. Normalize login URL hints to allowlisted HTTPS Claude
and Anthropic hosts. Consume JSON escapes during redaction. Preserve
failed sandbox cleanup state across retries and restarts, and prevent
deletion from severing the lifecycle context of a live reusable sandbox.
**Alternatives considered:** Keep raw diagnostics in Test checks or
trust login URL text from the sandbox. Both choices increase information
exposure. Keep separate probe behavior in the ACP lane. That choice
would leave the two Test lanes inconsistent.
## What Changed
- Surface the sandbox login panel on both Test lanes.
- Reconcile the bundled Daytona plugin manifest so
`supportsSetupTokenLogin` reaches the UI capability gate.
- Prepare the ACP Test lane with the same probe runtime as the CLI Test
lane.
- Add the `claude_acp_login_probe_unavailable` warning when the ACP
probe cannot run.
- Send raw sandbox diagnostics only to redacted server logs.
- Keep Test checks on fixed public messages in the ACP, managed-config,
and CLI paths.
- Normalize login URL hints to allowlisted HTTPS Claude and Anthropic
hosts.
- Redact JSON and escaped-JSON secret values, including escaped quotes
and backslashes.
- Preserve orphan cleanup records across provider failures, restarts,
and unavailable plugins.
- Atomically block environment deletion while a live reusable sandbox
lease still depends on it.
- Verify pending cleanup destroys plugin sandboxes with the provider
configuration recorded on the lease, even after the current environment
configuration changes.
## Verification
- Head under review: `506b7fa2d83c36bfa5fd722ee9d95b0c7431c241`.
- Focused environment route/service/runtime coverage passes: 196 tests
across 3 files.
- `pnpm -r typecheck` passes.
- `pnpm build` passes.
- The full Vitest run completed with 4,754 passing and 28 failing tests.
All 23 source-test failures reproduce unchanged on parent head
`58cfe61a33191ce03d965d65085d26064b4888ba`; the other 5 are duplicate
executions from stale `server/dist` output. The failures are unrelated
macOS path/listener and scheduler-fixture failures, so there is no new
bad commit for bisect to localize.
- All required CI checks pass for the current head, including build,
typecheck/release registry, all server and workspace shards, serialized
server suites, canary, and e2e.
- A fresh Greptile review for `506b7fa2d83c36bfa5fd722ee9d95b0c7431c241`
reports 5/5, “safe to merge,” with no blocking failure remaining.
## Risks
- A probe or redaction change could hide useful server diagnostics.
- An allowlist change could reject a valid Claude login URL.
- Cleanup recovery changes could affect provider teardown ordering.
- An environment with a live reusable sandbox can no longer be deleted
until the owning issue or execution workspace completes teardown.
- The implementation keeps public Test messages fixed and sends detail
to redacted server logs.
## Model Used
OpenAI GPT-5 via Codex — exact model ID: GPT-5; tool use and code
execution enabled; extended reasoning enabled. The implementation author
used AI-assisted development.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` /
`Closes: #` / `Refs: #` OR (b) described the issue in-PR following the
relevant issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and documented the result
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation or confirmed no separate
documentation change is needed
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - The agent configuration UI controls each agent and its allowed
secrets.
> - The environment variable editor already has a secret selector with
search and folder navigation.
> - The secret access editor used a basic list and made large secret
stores hard to use.
> - The secret access controls also occupied the main Configuration tab.
> - This pull request reuses the rich selector and moves secret access
to a dedicated Secrets tab.
> - The benefit is one consistent secret selection workflow with clearer
agent configuration navigation.
## Linked Issues or Issue Description
**What existing behavior does this improve?**
The agent detail configuration view and its secret access editor.
**Subsystem affected**
`ui/` — React and Vite board UI.
**Current behavior**
The secret access editor uses a basic select control. It does not
provide the search and folder navigation available in the environment
variable editor. The editor also appears inside the Configuration tab.
**Proposed behavior**
The secret access editor uses the shared secret picker. Users can search
secrets and browse slash-delimited folders. Agent details provide a
dedicated Secrets tab for this editor.
**Reason and benefit**
Large secret stores are slow to scan in a flat list. Reusing one
selector reduces UI differences and makes scoped secret access easier to
manage.
**Breaking changes**
None. The API and saved secret access data do not change.
## What Changed
- Reused the environment variable secret picker in the agent secret
access editor.
- Preserved secret version selection and the create-secret action,
including nested-popover focus handling.
- Added a route-backed Secrets tab to agent details and removed secret
access controls from Configuration.
- Guarded unsaved configuration across tab, link, browser-history, and
action-triggered navigation.
- Rechecked dirty state when navigation-producing agent actions finish,
covering edits made while a request is pending.
- Added component, page, and Storybook coverage for the workflow.
## Verification
- `pnpm --filter @paperclipai/ui exec vitest run
src/components/AgentActionButtons.test.tsx
src/components/AgentConfigForm.render.test.tsx
src/components/AgentSecretAccessEditor.test.tsx
src/components/environment-variables-editor/EnvironmentVariablesEditor.test.tsx
src/pages/AgentDetail.progress.test.ts` — 82 tests passed.
- `pnpm --filter @paperclipai/ui typecheck`
- `pnpm check:token-gates`
- All GitHub PR checks passed on `5209c5b787`, including build, canary,
general and serialized tests, and all three e2e shards.
- Greptile completed at 5/5 with zero unresolved review threads.
## Risks
- Low risk. The API and persisted binding format are unchanged; this
changes agent configuration navigation and secret selection UI.
- Dirty-state guards now cover direct navigation, Back/Forward history,
and navigation-producing agent actions, including pending-request races.
- Tests cover tab separation, secret access updates, search, folder
navigation, focus restoration, and navigation rejection.
- No documentation change is required because commands, contracts, and
setup steps do not change.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex with GPT-5. This runtime did not expose a more specific
model ID or context window. The model used agentic reasoning, repository
tools, and code execution.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip helps people manage AI agents for work.
> - Agent adapters connect Paperclip to tools such as the Codex command
line tool.
> - A sandboxed Codex agent may start without a credential.
> - The operator needs a safe sign-in flow that does not expose
credentials to the shared package or the sandbox.
> - This pull request adds a company-scoped device-login flow with a
temporary Daytona sandbox.
> - The flow promotes the credential only after readiness checks pass
and removes the temporary sandbox after use.
> - The result lets an operator sign in to a sandboxed Codex agent from
the agent form.
## Linked Issues or Issue Description
**Subsystem affected**
Cross-cutting (multiple of the above)
**Problem or motivation**
A Codex adapter that runs in a sandbox cannot authenticate when the
company has no pre-provisioned Codex credential.
**Proposed solution**
Add a company-scoped device-login session. Start a temporary sandbox,
run `codex login --device-auth`, stream the code and URL, verify
readiness, promote the credential, and delete the sandbox.
**Alternatives considered**
Pre-provisioning a credential does not support first-time sandbox login.
Keeping the credential in the login sandbox does not provide a durable
company credential.
**Roadmap alignment**
This supports the roadmap item for cloud and sandbox agents.
**Additional context**
The flow uses a five-minute cleanup reaper, compare-and-set status
changes, and a PostgreSQL advisory lock to protect promotion and
cleanup.
## What Changed
- Add the adapter login-session contract, database table, and migration.
- Add company-scoped server routes and a service for sandbox device
login.
- Add credential promotion, readiness checks, and cleanup after login.
- Add restart-safe cleanup for abandoned login sandboxes.
- Add sandbox login controls to the agent creation and edit forms.
- Keep device-login and vendor identifiers out of public shared and
adapter UI symbols.
## Verification
- `pnpm --filter @paperclipai/adapter-codex-local exec vitest run`
passed with 310 tests at the submitted commit.
- The server login route, service, and reaper tests passed with 45 tests
at the submitted commit.
- The agent form render tests passed with 26 tests at the submitted
commit.
- The public-symbol leak check passed at the submitted commit.
- A live Daytona sign-in flow still requires confirmation by a user with
a live sandbox.
## Risks
The migration adds a new company-scoped table. A promotion or cleanup
race could remove a credential or leave a sandbox active, so the service
uses claims, compare-and-set transitions, and an advisory lock. The live
Daytona flow needs operator confirmation because local tests do not
provide a real browser sign-in.
## Model Used
OpenAI Codex, GPT-5, tool use and code execution, extended reasoning.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` /
`Closes: #` / `Refs: #` OR (b) described the issue in-PR following the
relevant issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
<!-- Write all pull request text in Simplified Technical English
(ASD-STE100): short sentences, one instruction per sentence, simple
approved vocabulary, and the active voice. -->
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Environments give each agent run an execution target, and sandbox
providers (Daytona, E2B, Novita, exe.dev) run as plugin workers
> - A managed deployment provisions one platform-managed sandbox row
with no credential in config; the provider is documented to fall back to
its process env var (for example `DAYTONA_API_KEY`)
> - Plugin workers spawn with a scrubbed environment, so that fallback
never sees the host env var — probe and lease acquisition fail with
"require an API key in config or DAYTONA_API_KEY" even when the
deployment sets the var
> - Separately, the managed-sandbox-only mode hides local rows from
every list, but the instance Default picker renders a hardcoded
synthetic "Local" option that no filter touches
> - This pull request forwards each bundled provider's documented
credential env var to its own plugin worker, and gates the synthetic
Local option on the flag
> - The benefit is that the documented host-env credential fallback
works for plugin-backed providers, and managed-sandbox-only instances no
longer offer Local anywhere
## Linked Issues or Issue Description
**Subsystem affected**
Plugin worker environment construction
(`server/src/services/plugin-loader.ts`) and the environments UI
(instance Default picker, agent form inherited-environment label).
**Problem or motivation**
Two follow-ups to the managed-sandbox-only mode (#11200), both found on
a live managed deployment:
1. The deployment sets `DAYTONA_API_KEY` as a server env var and the
managed sandbox row omits `config.apiKey` by contract. "Test Connection"
fails with `Sandbox environment probe failed for provider "daytona".
Daytona sandbox environments require an API key in config or
DAYTONA_API_KEY.` A real agent run fails the same way at lease
acquisition. The cause: sandbox providers run as plugin workers, and
`buildPluginWorkerEnv` passes only model-provider keys and in-cluster
Kubernetes vars. The provider's own documented credential env var never
reaches the worker, so the in-plugin `process.env` fallback reads
nothing. The self-hosted path has the same gap: the Daytona plugin
README documents `DAYTONA_API_KEY` as a host-level fallback, and it does
not work today.
2. With `enableManagedSandboxOnly` on, the instance Default environment
picker still shows "Local". The server filters local *rows* out of the
list, and the client filter mirrors that for cached lists, but this
option is a hardcoded `<option value="">Local</option>` — not a list row
— so no filter removes it. Selecting it writes a null default, which run
selection then rejects fail-closed.
**Proposed solution**
Forward each bundled sandbox provider's documented credential env var
into its plugin worker, keyed by the manifest's declared
`environmentDrivers[].driverKey` so a worker only receives its own
provider's credential (daytona → `DAYTONA_API_KEY`, e2b → `E2B_API_KEY`,
exe-dev → `EXE_API_KEY`, novita → `NOVITA_API_KEY`). Keep the existing
gate: only plugins that declare `environment.drivers.register` receive
any passthrough. In the UI, render the synthetic Local option only when
managed-sandbox-only is off; under the flag show a disabled "Select
environment" placeholder only while no default is stamped yet, and stop
the agent form's inherited label from reading "Local".
**Alternatives considered**
Adding `DAYTONA_API_KEY` to the existing `ADAPTER_ENV_PASSTHROUGH` list
was rejected: that list goes to every environment-driver plugin, so each
provider would receive every other provider's credential. A manifest
schema field for declared credential env vars was rejected as heavier
than needed: the bundled providers are known, and the mapping lives next
to the two existing passthrough lists.
## What Changed
- `server/src/services/plugin-loader.ts`: new
`SANDBOX_PROVIDER_CREDENTIAL_ENV_PASSTHROUGH` map (driverKey →
documented credential env vars). `buildPluginWorkerEnv` reads the
manifest's `environmentDrivers` and forwards only the matching vars,
after the existing `environment.drivers.register` gate. Blank values
stay excluded.
- `server/src/__tests__/plugin-database.test.ts`: the daytona worker
receives `DAYTONA_API_KEY` and not another provider's key; a plugin
whose drivers have no mapping (kubernetes) receives no credential var.
- `ui/src/pages/CompanyEnvironments.tsx`: the Default picker's synthetic
Local option renders only when managed-sandbox-only is off. Under the
flag, a disabled "Select environment" placeholder renders only while the
default is unset.
- `ui/src/pages/CompanyEnvironments.test.tsx`: the Local option is
present by default and absent under the flag; saved non-local
environments stay selectable.
- `ui/src/components/AgentConfigForm.tsx`: the inherited-environment
label falls back to "Managed sandbox" instead of "Local" under the flag.
## Verification
- `server`: `npx vitest run src/__tests__/plugin-database.test.ts -t
buildPluginWorkerEnv` — 5 passed (3 existing, 2 new).
- `ui`: `npx vitest run src/pages/CompanyEnvironments.test.tsx` — 22
passed (2 new); `npx vitest run
src/components/AgentConfigForm.render.test.tsx` — 10 passed.
- `tsc --noEmit` clean in `server` and `ui`.
- Live managed deployment: confirmed the tenant service env carries
`DAYTONA_API_KEY` while the probe fails with the exact message above,
which pins the root cause to the worker env, not delivery.
## Risks
- The worker env grows by exactly one var per matching bundled provider,
only when the deployment sets it and only for plugins that declare a
matching environment driver. Plugins without a mapping see no change.
- Self-hosted behavioral shift is the fix itself: a host-level
`DAYTONA_API_KEY` (or E2B/EXE/NOVITA equivalent) now reaches the
provider as its README documents. Deployments that set the var but
expected it to stay inert had no working configuration to preserve — the
provider errored on every keyless probe and run.
- UI change is inert unless `enableManagedSandboxOnly` is on (default
false everywhere).
## Model Used
Claude Fable 5 (`claude-fable-5`) via Claude Code — extended thinking,
tool use, parallel read-only subagents for the two root-cause traces.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the control plane people use to manage AI-agent
companies.
> - Agents can encounter credentials during work.
> - Directly creating live secrets or bindings would bypass human
governance.
> - Proposal records must remain inert and separate from live secret
resolution until an authorized human approves them.
> - Approval must reuse the existing secret-create and protected
agent-config write paths.
> - This pull request adds the propose, review, approve, and reject
lifecycle.
> - The benefit is that agents can safely hand credentials into
Paperclip without exposing plaintext or gaining authority to activate
them.
## Linked Issues or Issue Description
Follow-on to #9921, which established run-bound agent secret access.
**Problem / motivation:**
Agents can receive credentials during work. There is no governed way for
them to propose a credential or binding without exposing plaintext in
work artifacts or immediately creating live access.
**Proposed solution:**
Store agent-authored proposals outside live secret tables. Encrypt each
proposed value and register exact-value redaction when Paperclip
receives it. Require an authorized human to approve or reject each
proposal. Approval executes through the normal write paths as the human
approver. Binding proposals can target only the proposer or its downward
reporting chain under the restrictive V1 policy.
**Alternatives considered:**
We rejected live secrets with a `proposed` status. That design would put
untrusted rows in resolver, list, and sync paths. It would also allow
uniqueness squatting. We rejected direct agent binding writes because a
binding is an agent-config write and must keep the existing human
permission gate.
**Roadmap alignment:**
This change extends the run-bound agent secret-access foundation in
#9921 with a governed proposal workflow.
## Security Verdict
Q0 SecEng verdict: **PASS-with-required-changes**. The review accepted
the separate proposal-table design and required the implementation to:
- fail closed unless both encryption and exact-value run redaction
registration succeed;
- scrub ciphertext idempotently on reject, withdraw, and expiry, with
audit-visible state;
- treat agent justification as hostile input and foreground action,
target, provenance, and approver permissions;
- snapshot and re-check the target agent plus reports-to chain at
approval to prevent org-chart laundering;
- make cascade approval atomic and fail closed if either secret creation
or binding authorization fails;
- deny low-trust, `skill_test`, `task_bridge`, and non-run-bound sources
consistently; and
- execute approval through the normal human secret/config write paths,
including protected-change gates.
Those requirements are implemented and covered by focused service,
route, and UI tests. Residual V1 risk remains the accepted 14-day
encrypted retention window. Proposal-time redaction also cannot clean a
value that leaked before the propose call.
## What Changed
- Added `company_secret_proposals`, migration `0207`, shared proposal
contracts, and a state-machine service for create, approve, reject,
withdraw, cascade, expiry, and ciphertext scrubbing.
- Added run-bound agent proposal routes and board review routes. The
routes derive provenance from authentication and enforce source
restrictions, company isolation, chain-of-command checks,
approval-as-approver, wake-on-resolution, and dual audit trails.
- Added durable per-run exact-value redaction registration so proposal
values remain redacted on later read surfaces.
- Added the Secrets **Proposals** tab and agent configuration **Proposed
access** rows. The UI shows fingerprint and length only. It also frames
agent justification as untrusted input, runs permission preflight,
supports approve and reject actions, and confirms cascades.
- Updated OpenAPI, agent skill guidance, API reference documentation,
and focused server and UI regression coverage.
- Rebased the branch onto current `master` and renumbered the proposal
migration after `0206`.
## QA Acceptance Results
Q5 QA verdict: **PASS — 9/9 acceptance criteria met**, with one Minor
non-blocking follow-up.
- **AC1:** proposed values never echo, never appear in live
lists/resolvers, and expose only fingerprint + length to board
reviewers.
- **AC2:** restrictive `self_and_reports` matrix passes: self/downward
allowed; upward/lateral denied.
- **AC3:** secret approval uses the normal create path, honors rename
overrides, records proposer/approver provenance, and scrubs ciphertext.
- **AC4:** approved bindings materialize and resolve through the target
agent's runtime list/fetch routes.
- **AC5:** pending-secret bindings require cascade; cascade succeeds
atomically and permission failures leave nothing applied.
- **AC6:** reject, withdraw, dependent rejection, and expiry paths scrub
ciphertext and preserve reasons/audit state.
- **AC7:** token/source and approver denial matrix passes through live
checks plus focused route tests.
- **AC8:** proposal lifecycle events and reused
`secret.created`/config-write events form the required dual audit trail;
origin-issue notification and wake are queued.
- **AC9:** both review surfaces render and execute correctly; UI
approval materializes the binding.
QA also confirmed zero plaintext occurrences for all exercised proposal
values in server logs. The single finding is that the company-level
`bindingTargetPolicy` toggle is not wired yet. V1 is hardcoded to the
restrictive `self_and_reports` policy. The matrix is correct and the
follow-up is tracked separately, so QA classified it as non-blocking.
## Verification
- Focused server proposal and redaction suite: 83 tests pass.
- Focused proposal review UI suite: 54 tests pass.
- Embedded-Postgres migration reapply test: 1 test passes with the
documented 30-second timeout.
- `pnpm --filter @paperclipai/db typecheck` passes, including migration
numbering and safety checks.
- `pnpm --filter @paperclipai/shared typecheck` passes.
- `pnpm --filter @paperclipai/ui typecheck` passes.
- `pnpm check:token-gates` passes with all gates clean.
- Q5 exercised the complete propose, review, approve, bind, and
runtime-resolve flow over real HTTP, JWT, and database paths. It
verified 9/9 acceptance criteria.
## Risks
- Proposal ciphertext is retained encrypted for up to 14 days while
pending. Terminal-state and expiry scrub paths reduce but do not remove
server-compromise risk during that window.
- The V1 target policy is restrictive but not yet company-configurable.
A separate follow-up owns that change.
- A new migration can require another renumber if another migration
lands before maintainers merge this pull request.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex coding agent. The exact runtime model ID and
context-window size are not exposed. The agent used reasoning,
repository editing, terminal execution, Paperclip API, and GitHub CLI
capabilities. Q3 UI work also records Claude Opus 4.8 assistance in its
commit trailers.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal or instance-local Paperclip issues
or links
- [x] My branch name describes the change and contains no internal
Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Agents execute in environments — local, SSH, or sandboxes — resolved
at run time as agent environment → instance default → local
> - The Configuration page has a Test button that probes the adapter
(working directory, command, a model call) in the environment it will
run in
> - But the Test sent only the agent's own environment id, with no
instance-default fallback, so agents relying on the instance default
were probed on the Paperclip host instead
> - A sandbox image carrying an extra CLI then fails the Test with
"command not found" even though every real run would resolve to the
sandbox and succeed — the Test lies about a working setup
> - This pull request mirrors the run-time resolution in the Test call
via a small shared helper with tests
> - The benefit is that the Test button reports the truth about where
the agent actually runs
## Linked Issues or Issue Description
No existing public issue — inline description following the bug report
template:
**What happened?**
With the instance default environment set to a sandbox (whose image
includes the adapter CLI) and an agent that leaves its environment unset
("use instance default"), the Configuration page's Test fails with
`command not found` for that CLI.
**Expected behavior**
The Test probes the environment a real run would use — here the
instance-default sandbox, where the CLI exists — and passes.
**Steps to reproduce**
1. Set the instance default environment to a sandbox whose image carries
an adapter CLI not installed on the Paperclip host (e.g. `grok`).
2. Create a `grok_local` agent without selecting an environment.
3. Press Test on the agent's Configuration page → `command not found`,
while a real heartbeat run resolves to the sandbox and works.
**Paperclip version or commit**
Reproduced on `sha-53bcf38-cloud`-era master; root-caused in
`ui/src/components/AgentConfigForm.tsx` (`environmentId =
currentDefaultEnvironmentId || null`) versus the server's
`resolveExecutionWorkspaceEnvironmentId` (agent → instance default →
local).
## What Changed
- New `ui/src/lib/adapter-test-environment.ts`:
`resolveAdapterTestEnvironmentId` — agent environment first, else
instance default, else null (host probe) — documented as the mirror of
the server's run-time resolution.
- `AgentConfigForm` uses it in the Test mutation. The raw agent
environment id is now sent even when it points at the local environment:
the server already resolves the driver and probes the host for local, so
explicit-local behavior is unchanged, and the test-environment route's
remote paths (SSH/sandbox lease + custom-image template) engage exactly
as they do for the fallback environment.
- Tests pin the fallback (agent wins; instance default when agent unset;
null when neither).
Deliberately untouched: the onboarding wizard's adapter test still sends
no environment — during onboarding an instance default frequently
doesn't exist yet, and changing that flow deserves its own look.
## Verification
- `vitest run` on the new helper suite plus both `AgentConfigForm`
suites — 19 tests pass; `tsc` clean in `ui/`.
- Root cause verified against a live deployment: an agent with
`default_environment_id = NULL`, instance default = sandbox environment;
the Test posted `environmentId: null` and probed the host (no `Probing
inside environment: …` check in the result), which lacks the CLI that
the sandbox image carries.
## Risks
- Low. The change only widens which environment the Test probes,
matching run-time reality. Sandbox-backed tests boot a throwaway sandbox
(existing route behavior — lease, custom-image template,
archive-on-release), so Tests for instance-default-sandbox agents now
take sandbox-boot time instead of failing fast and wrongly.
## Model Used
Claude Fable 5 (`claude-fable-5`, extended thinking, via Claude Code
with tool use and code execution); diagnosis included live inspection of
a deployed instance's agent/environment configuration.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
(helper doc-comment carries the rationale; no user-facing doc covers the
Test button)
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source control plane people use to manage
AI-agent companies.
> - Agents already receive selected company secrets through `env.*`
bindings at run launch, but environment injection is ambient,
long-lived, and not suitable for every secret consumer.
> - The existing binding and secret-access-event models already provide
company-scoped authorization and per-resolution audit seams.
> - Agents need an explicit way to discover only the secrets granted to
them and fetch a value on demand without exposing the wider company
catalog.
> - That capability must remain run-bound, preserve low-trust token
carve-outs, and make every value read visible in both security and
operator audit trails.
> - This pull request adds an `access.*` delivery namespace, two
run-bound agent routes, dual audit logging, documentation, and an
operator grants editor.
> - The benefit is least-privilege, revocable, auditable secret access
while preserving existing env injection behavior.
## Linked Issues or Issue Description
No pre-existing public issue. Related work:
- Refs #9797 — existing in-sheet agent access UI that this PR extends to
distinguish env and API delivery.
- Refs #9918 — complementary searchable-agent picker improvement for the
same secrets sheet.
- Refs #9530 — related company-wide metadata catalog proposal; this PR
intentionally exposes only the authenticated run's granted aliases and
values.
**Problem / motivation:** Agents can currently consume secrets only
through process environment injection. This keeps values resident for
the run, does not support on-demand consumers, and cannot provide a
discrete operator-visible activity event for each agent-initiated read.
**Proposed solution:** Treat `company_secret_bindings` as the source of
truth for agent secret grants. Keep `env.KEY` as env delivery and add
`access.ALIAS` for API-only delivery; an env binding also implies read
access because the value is already present in the agent process. Add
run-bound list/fetch endpoints that derive scope from the authenticated
heartbeat run and never accept caller-selected overlays.
**Alternatives considered:** A company-wide agent-readable catalog was
rejected for this value path because it increases reconnaissance and
does not prove a per-secret grant. Reusing the ephemeral
environment-probe resolver was rejected because it lacks binding
enforcement. Approval-gated reads and user-scoped secrets remain
deferred beyond v1.
**Roadmap alignment:** This extends the completed **Secrets Manager with
per-agent access** roadmap capability from launch-time env injection to
explicit run-bound API delivery without duplicating a separate planned
initiative.
## What Changed
- Added `access.*` agent binding validation and a dedicated run-bound
resolver that combines `secrets:read` authorization with binding-context
enforcement.
- Added `GET /api/agents/me/secrets` for minimal granted metadata and
`POST /api/agents/me/secrets/:key/value` for on-demand value fetches
with `Cache-Control: no-store`.
- Preserved the existing denials for low-trust review agents,
task-bridge credentials, and skill-test tokens; standard long-lived
agent API keys cannot call the run-bound routes.
- Added dual audit behavior: value attempts write `secret_access_events`
and `activity_log` (`secret.value.read`), while metadata listing writes
the lighter `secret.access.listed` activity event.
- Kept env compatibility: `env.*` remains injected at launch and also
implies API read for the same bound agent; `access.*` never becomes an
environment variable.
- Added the agent-settings **Secret access** editor plus
delivery-mode/alias surfacing on the Secrets page, with focused UI tests
and tokenized layout styles.
- Updated OpenAPI, shared types, agent-facing skill documentation, and
API reference documentation.
### UI Screenshots
P3 produced and reviewed three screenshots using mock data; images are
intentionally not committed to the repository:
- `secret-access-editor.png` — agent settings grant editor.
- `secret-access-light.png` — Secrets-page delivery surfacing in light
mode.
- `secret-access-dark.png` — Secrets-page delivery surfacing in dark
mode.
The source attachments are retained with the implementation task and
linked in the internal handoff; the public page publisher was
unavailable in the PR-prep runtime.
## Verification
- `pnpm exec vitest run
server/src/__tests__/agent-secrets-routes.test.ts
server/src/__tests__/secrets-service.test.ts
server/src/__tests__/secrets-routes.test.ts
ui/src/lib/secret-delivery.test.ts
ui/src/components/AgentSecretAccessEditor.test.tsx` — 5 files, 122 tests
passed.
- Security follow-up: `pnpm exec vitest run
server/src/__tests__/agent-secrets-routes.test.ts
server/src/__tests__/secrets-service.test.ts` — 2 files, 73 tests passed
after active-run and version-consistency fixes.
- Final-head CI: all feature, typecheck, build, e2e, security, and
review gates pass; `General tests (server (1/3))` remains red after one
rerun because unrelated `heartbeat-retry-scheduling.test.ts` cleanup
deletes `heartbeat_runs` before referenced `activity_log` rows.
- `pnpm --filter @paperclipai/server typecheck` — passed.
- `pnpm --filter @paperclipai/shared typecheck` — passed.
- `pnpm --filter @paperclipai/ui typecheck` — passed.
- `pnpm check:token-gates` — feature-local arbitrary-value violations
fixed; command still reports five unchanged `#9627` literals outside
this PR.
- End-to-end QA passed all eight acceptance criteria: grant/list, fetch,
dual audit, env-implies-read, denial matrix, revocation, UI rendering,
and env-injection regression. Evidence:
https://github.com/paperclipai/paperclip/pull/9921#issuecomment-5027455492
- Security review returned PASS-with-required-changes; the
implementation uses the required dedicated binding-enforcing resolver,
run-bound JWT restriction, run-derived overlays, minimal metadata, and a
resolver redaction-registration hook. Evidence:
https://github.com/paperclipai/paperclip/pull/9921#issuecomment-5027455382
## Risks
- A compromised agent can exfiltrate any secret explicitly granted to
it; explicit company-scoped/run-scoped grants, revocation, and audit
reduce but cannot remove that inherent capability risk.
- The resolver invokes a redaction-registration hook before returning
values, but the current route has no persistent cross-request per-run
redaction registry. Paperclip-owned later comments/events therefore
cannot yet guarantee automatic scrubbing of a deliberately copied
fetched value; QA classified this as non-blocking residual hardening.
- Audit-event insertion currently fails open if the security-event
insert itself fails; the operator activity event provides partial
redundancy, but a future hardening change should define fail-closed
behavior for value delivery.
- This PR overlaps `ui/src/pages/Secrets.tsx` with #9918 and may require
a straightforward rebase after that PR moves.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex, `gpt-5.3-codex`, with reasoning, repository tool use,
terminal execution, Paperclip API access, and GitHub CLI capabilities.
Context-window size is not exposed by the runtime.
- Anthropic Claude Opus 4.8 with 1M context and tool use assisted with
the UI implementation commit.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent configuration includes adapter-specific model settings and a
built-in adapter test action so operators can verify runtime
configuration before saving changes.
> - Cody/Codex-style local adapters can use an adapter default model
when the user clears the explicit model field.
> - The adapter test path still passed an object containing `model:
undefined` in some create/edit flows, which is different from omitting
the model and can break default-model behavior.
> - The previous fix was reverted because it also included an unrelated
skill documentation edit.
> - This pull request reapplies only the UI default-model test-config
fix, with no doc or skill changes.
> - The benefit is that testing Cody/Codex adapter settings with the
default model follows the same contract as saving default model
settings: no explicit model key is sent.
## Linked Issues or Issue Description
Bug report:
- Summary: Testing a Cody/Codex local agent after selecting the default
model could send an adapter config with an undefined model value instead
of omitting the model key.
- Expected behavior: Clearing the model to use the adapter default
should test with `adapterConfig: {}` unless another model is explicitly
selected.
- Actual behavior: The UI test-config path could preserve `model:
undefined`, causing the adapter test to fail instead of exercising the
default model.
- Related PRs: Reapplies the UI-only portion of #9361 after #9363
reverted the original PR.
## What Changed
- Exported and reused `omitUndefinedEntries` so adapter test config
payloads drop undefined adapter config entries before calling the test
endpoint.
- Hardened the current model display value so create-mode values that
are nullish or non-string do not crash the model selector/test flow.
- Added render coverage for editing a Codex agent back to the default
model and for testing a create form with the default model.
## Verification
- `pnpm exec vitest run
ui/src/components/AgentConfigForm.render.test.tsx`
- `pnpm check:token-gates`
- Confirmed `git diff origin/master --name-only` contains only:
- `ui/src/components/AgentConfigForm.render.test.tsx`
- `ui/src/components/AgentConfigForm.tsx`
- `ui/src/lib/agent-config-patch.ts`
## Risks
Low risk. The change only removes `undefined` adapter config entries
from the UI adapter-test payload and adds focused render coverage.
Explicit model values and other adapter config fields are preserved.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex, GPT-5 coding agent, tool-use enabled. Context window size
not exposed in this runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - The board UI lets operators create and edit agent adapter
configuration, including a primary model field and an adapter test
action.
> - For Cody and similar adapter forms, selecting the default model
means the model value is intentionally unset so the adapter can use its
default.
> - The adapter test path still allowed `model: undefined` to survive in
the generated adapter config, which could send an invalid test payload
instead of omitting the field.
> - This pull request normalizes create/edit adapter test config so
default-model selections omit `model` entirely.
> - The benefit is that testing an agent configured to use the adapter
default model exercises the same clean config shape that should be saved
and run.
## Linked Issues or Issue Description
No public GitHub issue was found for this local UI bug, so the problem
is described inline.
Bug description:
- What happened: using the adapter test action after choosing the
default model could include `model: undefined` in adapter config and
surface a UI/runtime error instead of testing with the adapter default.
- Expected behavior: choosing the default model should omit the `model`
field from adapter config so the adapter default is used.
- Steps to reproduce: edit a Codex/Cody-style agent with a concrete
model, switch the model selector to Default, then run the adapter Test
action.
- Paperclip version/commit: current `master` before this PR.
- Deployment mode: board UI, deployment-mode independent.
## What Changed
- Sanitized adapter test config assembly so undefined adapter config
entries are omitted before the test request is sent.
- Made create-mode current model display resilient when the model is
unset for adapter defaults.
- Added regression coverage for editing an existing agent from a
concrete model back to Default and testing it.
- Added regression coverage for create-mode testing with an
unset/default model.
- Hardened the developer skill wording used by the existing server skill
utility contract test.
## Verification
- `pnpm --filter @paperclipai/ui exec vitest run
src/components/AgentConfigForm.render.test.tsx`
- `pnpm exec vitest run
server/src/__tests__/paperclip-skill-utils.test.ts`
- GitHub PR checks on this branch are green, including Typecheck +
Release Registry, Build, General tests, e2e, verify, security scans, and
Greptile Review.
## Risks
- Low risk: this only removes undefined values from adapter test config
payloads, which aligns with the existing persisted patch behavior.
- Low risk: default-model display now treats unset create-mode model
values as an empty string.
- No database, API schema, migration, auth, or adapter runtime contract
changes.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected - check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex, GPT-5 coding agent, tool-enabled software-engineering
session. Exact context window size was not exposed by the runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Cody <cody@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Operators need first-party agent capabilities for repeatable company
work, not just manually created one-off agents.
> - Built-in agents need to behave like normal company-scoped agents
while preserving approval gates, permissions, budgets, and audit trails.
> - Reflection and coaching work also needs bundled instructions, skill
content, and a routine so the feature can be installed and reset
predictably.
> - The API, database, UI, portability, and tests all need to agree on
the built-in lifecycle from not provisioned through setup, approval,
ready, paused, and reset.
> - This pull request adds built-in agent provisioning and the
Reflection Coach bundle end-to-end.
> - The benefit is a safer first-party path for Paperclip-managed agents
without bypassing the same governance model used for operator-created
agents.
## Linked Issues or Issue Description
No public GitHub issue was found for this exact built-in agent and
Reflection Coach bundle work.
Problem/motivation:
- Paperclip did not have a first-party built-in agent lifecycle for
product-owned agents.
- Bundled agent resources such as default instructions, skills, and
routines needed managed ownership and reset semantics.
- Approval-gated companies needed built-in setup to preserve requested
adapter, budget, manager, and permission state through board approval.
- The board UI needed clear built-in badges, setup affordances,
readiness state, and bundle status without exposing secrets.
Proposed solution:
- Add a company-scoped built-in agent registry,
provisioning/reset/reconcile/status APIs, and Reflection Coach bundled
resources.
- Track bundled managed resources in the database with idempotent
migration behavior.
- Reuse existing agent approval, authorization, budget, and activity-log
paths instead of creating a bypass.
- Add UI setup, badges, gates, bundle panels, and route coverage for
built-in agents.
Duplicate search:
- Searched GitHub PRs for `built-in agents Reflection Coach
repo:paperclipai/paperclip`; only this PR was returned.
- Searched GitHub issues for the same query; no public issues were
returned.
## What Changed
- Added built-in agent definitions, lifecycle state derivation,
provisioning, reset, reconcile, status, and routine-control routes.
- Added the `built_in_managed_resources` migration and schema exports
for bundled instructions, skill, and routine ownership.
- Added the Reflection Coach built-in bundle with default instructions,
skill catalog content, routine template, default permissions, and
managed-resource drift handling.
- Added approval-aware provisioning behavior that preserves requested
adapter config, budgets, manager assignment, and built-in permissions
through hire approval.
- Added authorization and mutation gates for built-in agent and skill
changes, including consented Reflection Coach change paths.
- Added UI surfaces for built-in agent setup, roster/detail badges,
readiness gates, bundle status, routine controls, and route filtering.
- Added company import/export and validator coverage for built-in
managed resources and low-trust/red-team presets.
- Addressed Greptile follow-ups for pending approval reconciliation,
consent-gate error propagation, config-read authorization fallback,
approval-path manager preservation, and non-model adapter provisioning.
## Verification
Local verification:
- `git diff --check public/master..HEAD` passed.
- `pnpm check:token-gates` passed with all gates clean.
- `pnpm exec vitest run
ui/src/components/ConfigureBuiltInAgentModal.test.tsx` passed: 1 file, 4
tests.
- `pnpm exec vitest run ui/src/components/EntityRow.test.tsx
ui/src/pages/Agents.test.tsx ui/src/components/BuiltInAgentGate.test.tsx
ui/src/components/ConfigureBuiltInAgentModal.test.tsx
ui/src/components/BuiltInBundlePanel.test.tsx
ui/src/pages/InstanceExperimentalSettings.test.tsx
ui/src/pages/Routines.test.tsx` passed: 7 files, 64 tests.
- `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/built-in-agents.test.ts
src/__tests__/authorization-service.test.ts
src/__tests__/company-skills-routes.test.ts` passed: 3 files, 91 tests.
- `pnpm --filter @paperclipai/db check:migrations` passed.
- `pnpm -r typecheck` passed after the rebase; `pnpm --filter ui
typecheck` passed after the final UI review fix.
Remote verification on latest head
`1c61f693a4ec881d739022b0e75a8ca8bf8c2cd8`:
- Merge state: `CLEAN`.
- Greptile: `5/5`, zero unresolved Greptile threads.
- PR check rollup: all checks successful, neutral, or skipped as
expected.
- Passing gates include Build, Typecheck + Release Registry, all server
shards, all workspace shards, all serialized server suites, e2e, Canary
Dry Run, policy, review, verify, Socket, Superagent, and Snyk.
## Risks
- This adds a new managed-resource table and migration; the migration
uses idempotent create/add/index guards and passed migration safety
checks.
- Built-in agent provisioning touches approval and authorization paths;
tests cover pending approval preservation, stale retry rejection,
consent gates, and config-read fallback behavior.
- Reflection Coach creates managed instructions, skill, and routine
resources; drift/reset behavior is covered by service tests and redacted
API responses.
- Non-model adapter setup now provisions a `needs_setup` built-in row
before command/endpoint fields are complete; this matches the server
lifecycle and is covered by the setup modal regression test.
## Model Used
OpenAI Codex coding agent based on GPT-5. Exact hosted model ID,
context-window size, and reasoning-mode labels are not exposed in this
runtime; tool use, shell execution, GitHub CLI/API access, and local
code editing were enabled.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Adapter packages are the bridge between the control plane and local
agent harnesses such as Claude Code, Codex, and Gemini CLI.
> - ACP support was concentrated in a separate `acpx_local` adapter,
which made ACP feel like a separate agent choice instead of an execution
capability of the harness adapters.
> - Claude, Codex, and Gemini now have ACP-capable harnesses, so the
native adapter should own ACP selection, fallback, config, transcript
parsing, and environment diagnostics.
> - The standalone ACPX adapter still needs a compatibility path for
existing rows, but it should not be offered as an active adapter for new
agents.
> - This pull request moves the shared ACP runtime into
`@paperclipai/acpx-engine`, wires Claude/Codex/Gemini local adapters to
prefer ACP when prerequisites are available, and retires `acpx_local` to
a tombstone.
> - The benefit is one adapter per harness, richer ACP transcripts by
default where possible, and a migration path for existing Claude/Codex
ACPX agents.
## Linked Issues or Issue Description
Closes#5932 — the broken default `acpx_local` Claude path is replaced
by native `claude_local` ACP support, existing Claude/Codex ACPX rows
migrate to native adapters, and new agents no longer choose the
standalone ACPX adapter.
Refs #4893 — original merged ACPX local adapter runtime that this PR
replaces with native per-harness ACP engines.
Refs #6590 — prior ACPX-Claude seamlessness work folded into the new
native Claude ACP path.
Refs #197 — related open generic ACP/Kiro adapter work; this PR does not
close it because Kiro/custom generic ACP remains a separate adapter
decision.
Refs #7018 — related Kimi-specific `acpx_local` shell failure; this PR
retires the built-in standalone adapter but does not add a native Kimi
adapter.
Refs #8864 — related ACPX prompt/API guidance PR; this PR moves runtime
guidance into the shared/native ACP engine path instead of the old
standalone adapter.
Refs #8881 — related `acpx_local` POSIX shell failure from the old
`acpx` pin; this PR updates ACP dependencies but does not claim
custom/OMP ACP support as a first-class native adapter.
Refs #8964 — related open `acpx_local` stderr cleanup PR; this PR makes
the old runtime path obsolete for new agents but keeps it as a
non-closing reference.
Problem description:
- The standalone `acpx_local` adapter duplicates Claude/Codex agent
choices that already have first-class local adapters.
- ACP should be an execution engine capability of each harness adapter
when the underlying harness supports ACP.
- Existing `acpx_local` agents should either migrate to native harness
adapters or fail with an explicit retirement message instead of silently
falling back to the process adapter.
## What Changed
- Added `@paperclipai/acpx-engine` as the shared ACP execution,
session-codec, CLI formatter, and UI parser package.
- Wired `claude_local`, `codex_local`, and `gemini_local` to auto-select
ACP by default when prerequisites pass, with `engine=cli` opt-out and
`engine=acp` strict mode.
- Added ACP config schema/UI fields, environment checks, session-codec
preservation, transcript parsing, and adapter capability metadata for
the native adapters.
- Retired `acpx_local` to a server tombstone, removed its
UI/package/runtime image surface, and added a migration for existing
Claude/Codex ACPX agents.
- Updated package manifests, lockfile, release tooling, docs, Kubernetes
sandbox defaults, and tests.
## Verification
- `corepack pnpm --filter @paperclipai/acpx-engine typecheck`
- `corepack pnpm --filter @paperclipai/adapter-claude-local typecheck`
- `corepack pnpm --filter @paperclipai/adapter-codex-local typecheck`
- `corepack pnpm --filter @paperclipai/adapter-gemini-local typecheck`
- `corepack pnpm --filter @paperclipai/acpx-engine exec vitest run`
- `corepack pnpm --filter @paperclipai/adapter-claude-local exec vitest
run src/server/acp.test.ts src/server/execute.acp-fallback.test.ts
src/ui/build-config.test.ts`
- `corepack pnpm --filter @paperclipai/adapter-codex-local exec vitest
run src/server/acp.test.ts src/ui/build-config.test.ts`
- `corepack pnpm --filter @paperclipai/adapter-gemini-local exec vitest
run src/server/acp.test.ts src/ui/build-config.test.ts
src/ui/parse-stdout.test.ts`
- `corepack pnpm --filter @paperclipai/plugin-sdk ensure-build-deps &&
corepack pnpm --filter @paperclipai/server exec tsc --noEmit`
- `corepack pnpm --filter @paperclipai/server exec vitest run
src/__tests__/adapter-routes.test.ts
src/__tests__/adapter-session-codecs.test.ts
src/__tests__/adapter-models.test.ts`
- `corepack pnpm --filter @paperclipai/ui typecheck`
- `corepack pnpm --filter @paperclipai/ui exec vitest run
src/adapters/metadata.test.ts
src/adapters/adapter-display-registry.test.ts
src/components/AgentConfigForm.test.ts
src/components/AgentConfigForm.render.test.tsx
src/components/transcript/RunTranscriptView.test.tsx`
- `node --test scripts/bootstrap-npm-package.test.mjs
scripts/release-package-map.test.mjs
scripts/verify-release-registry-state.test.mjs`
Note: the server typecheck script calls `pnpm` internally; this dev
shell exposes pnpm through Corepack only, so I ran the two script steps
manually with `corepack pnpm`.
## Risks
- Migration changes existing `acpx_local` Claude/Codex agents to native
adapter types and clears old ACPX task sessions/runtime state.
- Custom ACP commands remain on the retired tombstone and will need a
separate future adapter/plugin path.
- ACP auto-selection depends on local Node and ACP server command
prerequisites; remote and unsupported environments fall back to CLI
unless `engine=acp` is explicit.
- `@paperclipai/acpx-engine` is a new public package and needs npm
trusted-publishing bootstrap before release automation can publish it.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI GPT-5 via Codex coding agent. Exact hosted model build and
context-window size are not exposed in this runtime. Tool use included
shell execution, repository editing, GitHub CLI operations, and local
test/typecheck execution.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Its UI is governed by a design system (`DESIGN.md` + the token layer
in `ui/src/index.css`, merged in #9134) whose first principle is "one
way to say each thing" — one Card, one Badge, one nav row
> - After the token extraction landed, ~35 files still hand-rolled card
containers, ~45 files hand-rolled pill spans, the sidebar agents section
duplicated the nav-row chrome, and the inbox and tasks lists rendered
the same task rows two subtly different ways
> - Each divergence is a place where a future design change (radius,
hover language, status vocabulary) silently misses surfaces, defeating
the "edit tokens + run checks" model the design system exists for
> - This pull request converges those surfaces onto the shared
primitives, codifies the radius scale as the modern multiplicative
shadcn ladder, and unifies the row/hover/tree-guide language across the
inbox and tasks lists — every visible delta was human-reviewed
screen-by-screen against a live instance across nine feedback rounds
> - The benefit is that the app's look is now steerable from single
knobs (one `--radius` anchor, one Card, one Badge, one row component),
and the visual regression suite covers the result (514 snapshots
including a new AgentDetail page story)
## Linked Issues or Issue Description
No public GitHub issue exists for this work; describing per the feature
template:
- **Problem**: after the design-token foundation (#9134),
component-level drift remained — hand-rolled cards/pills, duplicated
sidebar row chrome, and two different renderings of task rows (inbox vs
tasks list) meant design changes had to be applied per-surface and
frequently missed spots (e.g. status glyphs rendered 16px in the inbox
but 20px in the tasks list because a slot override silently beat the
component default).
- **Proposed behavior**: all card-shaped containers render via `Card`,
all label pills via `Badge`, sidebar rows via `SidebarNavItem`, and both
task-list surfaces via one `IssueRow` configuration; the radius scale is
a single multiplicative ladder anchored at `--radius: 0.5rem`.
- **Alternatives considered**: converting interactive
`<button>`/`<Link>` cards to `Card` divs (rejected — breaks semantics;
documented inline with `design-allow` comments instead); keeping the
legacy additive radius ladder (rejected in favor of the standard shadcn
multiplicative mapping).
## What Changed
- `Card` adoption across ~35 files (settings pages, auth/board flows,
dashboards, list containers, KPI tiles); non-adoptable sites
(interactive cards, `<li>` rows, class-string props, chart tooltip)
carry documented `design-allow(card-pattern)` comments
- `Card` gains an `interactive` prop — one quiet hover affordance for
clickable cards (cursor, border darken, shadow lift, focus ring),
applied to skills tiles, artifact cards, and the company selector; cards
carry no resting shadow
- `Badge` adoption for 113 hand-rolled pill spans across ~45 files;
`PropertyChip` wraps `Badge` internally; `StatusBadge`, external-object
chips, and match chips stay bespoke by documented decision (WCAG-tuned
status mechanics)
- Radius ladder becomes the multiplicative shadcn mapping
(`sm/md/lg/xl/2xl/3xl/4xl = 0.6/0.8/1.0/1.4/1.8/2.2/2.6 × --radius`,
anchor `0.5rem`); every card surface unifies on `rounded-lg`; the
orphaned 8px literal token is deleted
- Sidebar: agent rows render via `SidebarNavItem` (new additive props:
`iconNode`, `active`, `trailing`, `liveAccessory`); live dots use
`--status-agent-running`; one row rhythm and inset rounded pill
highlight; right-aligned trailing badges; every labeled section is
collapsible
- Inbox + tasks lists unified: md status glyphs, `accent/50` rounded row
hovers, vertical tree guides under parent rows (opaque underlay so
dark-mode translucent borders don't stack), no horizontal dividers under
expanded parents; the swipe-to-archive reveal layer shows only
mid-swipe; board toggle uses the `SquareKanban` glyph
- Kanban: every column carries a status-hued tint; lanes default
expanded (including empty); compact mode collapses empty lanes to
labeled rails (fixes a clipped, label-less empty-column state)
- Storybook: new AgentDetail page story (realistic fixtures, light+dark)
joins the visual suite; suite captures with `reducedMotion: 'reduce'`
and the ux-lab reasoning ticker honors `prefers-reduced-motion`; a stale
lexical alias in `storybook/main.ts` is fixed (build was broken since
the lexical 0.46 bump)
- Keyboard navigation, from live review of the unified lists: inbox
navigation keys work on every tab (archive keys stay scoped to the
archivable tab); keyboard-driven scrolling no longer hands the selection
to whatever row lands under the stationary cursor (hover selects only
after real pointer movement); the tasks list view gains the same j/k /
arrows / Enter selection model as the inbox; and the `g` then `i`
go-to-inbox chord works app-wide instead of only on the issue detail
page
- Token gates restored to 3/3 CLEAN (tokenized a post-#9134 regression
in the recovery card); decisions recorded in
`doc/design/DECISION-SHEET.md` and `doc/design/COMPONENT-INVENTORY.md`
(investigation verdicts: FileTree vs WorkspaceFileBrowser and the four
entity pickers stay separate — evidence included)
## Verification
- `pnpm check:token-gates` → 3/3 CLEAN
- `pnpm typecheck` → green (all packages)
- `cd ui && npx vitest run` → 2106/2106 (assertions updated in lockstep
where they documented superseded decisions; new tests for the global
go-to-inbox chord)
- `pnpm --filter @paperclipai/ui build` → exit 0
- `pnpm build-storybook` → succeeds (also fixes the lexical-alias break
on master)
- Visual regression: 514-snapshot Playwright suite green against the
updated baseline (zero diffs from the keyboard-navigation round — those
changes are purely behavioral). Note: baselines live outside git per the
suite design and the baseline-manifest archive is not yet published, so
CI cannot run this suite — it was run locally throughout; every visible
delta was reviewed screen-by-screen in a live instance across nine
review rounds. Review evidence (before/after triplets) intentionally
kept out of the repo for size; available on request.
- Manual: exercised dashboard, tasks (list + board), inbox, agents,
skills, costs, settings, and artifact surfaces in light and dark themes
## Risks
- Wide but shallow visual surface: most changes are class-string
substitutions with behavior preserved (props, handlers, roles, test
ids). The riskiest areas — dnd-kit card refs (React 19 ref-as-prop),
inbox swipe-to-archive, and sidebar overlays — are covered by existing
unit tests (all green) and were manually exercised.
- Intentional visual deltas (rounded cards, tinted kanban columns, md
status glyphs, unified hovers) are design decisions recorded in
`doc/design/DECISION-SHEET.md`; each maps to a re-baselined snapshot set
locally.
- The visual suite cannot yet run in CI (unpublished baseline archive —
pre-existing condition from #9134); until that lands, snapshot coverage
is local-only.
## Model Used
- Claude Fable 5 (`claude-fable-5`, Anthropic) via Claude Code — agentic
coding with tool use (file editing, test execution, Playwright
screenshot verification); extended thinking enabled.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green (will confirm once CI runs)
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
(pending first review)
- [x] I will address all Greptile and reviewer comments before
requesting merge
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Its UI is the operator's daily surface: task lists, boards, budgets,
agent status — all built on shadcn components and Tailwind
> - Visual values (colors, spacing, type sizes, radii) were hardcoded at
~1,600 call sites: the same "small gray label" was 9/10/11px depending
on the file, charts disagreed with chips about status colors, two
toggle-switch implementations coexisted in two greens, and there was no
visual regression coverage
> - This made the UI drift-prone and made any restyle a
hundreds-of-files project, which discourages design iteration
> - This pull request extracts visual values into a single token layer
in `ui/src/index.css`, adds a Storybook visual regression suite backed
by external immutable baseline archives, and then applies a deliberate
retune reviewed change-by-change on screenshot diffs
> - The benefit is that Paperclip's look becomes a config surface:
retheming is a token edit reviewed as a snapshot diff, drift is blocked
by a token gate, and future UI PRs can prove exactly what changed
visually without committing hundreds of PNGs
## Linked Issues or Issue Description
No existing public issue covers this work (searched "design tokens",
"visual regression", "design system" across issues and PRs). Related in
spirit: Refs #8982 (theming a hardcoded panel — a one-off instance of
the same problem class this PR addresses systematically).
**Problem (feature-request form):** UI visual values are hardcoded per
call site with no source of truth and no regression coverage;
consistency depends on reviewer memory, and restyling requires mass file
edits.
**Proposed solution (this PR):** a single token layer + enforcement gate
+ externally stored visual snapshot suite, then an intentional restyle
on top of that foundation.
## What Changed
- **Token extraction (zero visual change, machine-verified during
development):** committed codemods (`scripts/codemod-*.mjs`) moved
~1,600 hardcoded color/type/spacing/radius/shadow/misc values into named
tokens in a non-inline `:root` block of `ui/src/index.css`.
- **Visual regression suite:** `pnpm test:storybook-visual` covers 255
stories × light/dark = 510 Playwright screenshots at `maxDiffPixels: 0`,
plus new primitive-coverage stories and deterministic-render fixes.
- **External visual baselines:** committed PNG snapshots were removed.
`tests/storybook-visual/baseline-manifest.json` pins an immutable
archive URL/hash/size/count, and `scripts/storybook-visual-baseline.mjs`
handles `download`, `verify`, `pack`, and trusted maintainer `upload`
flows.
- **Opt-in visual CI artifacts:** added a `Storybook Visual` workflow
that runs on manual dispatch or PRs labeled `storybook-visual`,
downloads/verifies the baseline, runs Playwright, and uploads Playwright
report/test-result artifacts for review. Normal PR runs do not mutate
baseline objects.
- **Token gate:** `pnpm check:token-gates` — zero hex literals, zero
arbitrary bracket values, zero raw font-sizes in `ui/src/components/**`
and `ui/src/pages/**`, with a documented inline allowlist for legitimate
opt-outs.
- **Theme retune (intentional, snapshot-reviewed):** new base theme
values; radius ladder derived from a single `--radius` knob; micro-type
cluster collapsed to a named ladder (`--text-nano/micro/compact` +
Tailwind `text-xs`/`text-sm`); letter-spacing collapsed to named steps.
- **One status-color vocabulary:** charts, quota/budget bar fills,
RUNNING/live chips, and liveness indicators all use the canonical
`--status-*` hues. Light-mode legibility fixes for red alert surfaces
that used dark-tuned text classes.
- **One switch:** `ToggleSwitch` restyled to the registry capsule form,
second hand-rolled implementation removed, and all call sites unified.
- **Docs:** `DESIGN.md` is the design contract; `doc/design/` holds
audit reports, decision logs, and updated guidance for external baseline
review/update workflows.
- Dead code removed (`agentStatusBadge` duplicate map), byte-identical
contrast constants consolidated, semantic renames
(`--project-seed`/`--project-none`, `--liveness-blue`).
## Verification
- `pnpm check:token-gates` — 3/3 gates CLEAN during the design-system
run
- `pnpm typecheck` && `pnpm --filter @paperclipai/ui build` — green
during the design-system run
- `node --test scripts/__tests__/storybook-visual-baseline.test.mjs` —
pass after external-baseline rework
- `pnpm exec tsc --noEmit --pretty false --module NodeNext
--moduleResolution NodeNext --target ES2022 --types
node,@playwright/test tests/storybook-visual/playwright.config.ts
tests/storybook-visual/storybook-visual.spec.ts` — pass after
external-baseline rework
- `git diff --check origin/pr/9134..HEAD` — pass after external-baseline
rework
- `find tests/storybook-visual -type f -name '*.png' -print | wc -l` —
`0`
- `node scripts/storybook-visual-baseline.mjs verify` — intentionally
fails closed until the first trusted maintainer publishes the baseline
archive and updates `baseline-manifest.json`
## Risks
- **Large but shallow:** the PR still touches many UI files due to
mechanical token extraction and retune work, but committed PNG snapshot
churn has been removed from the branch.
- **Baseline publication required before the visual suite can pass in
clean clones:** the manifest currently has placeholder archive metadata.
A trusted maintainer must publish the first immutable archive, then
update `baseline-manifest.json`.
- **Rendering platform variance:** the external baseline should be
captured in the documented Linux/Chromium environment. Future CI runs
verify against the pinned archive and fail closed on checksum/count
mismatch.
- **Visual CI is opt-in while stabilizing:** add the `storybook-visual`
label or dispatch the workflow manually to produce downloadable
Playwright report/test-result artifacts.
- **Scheduled follow-ups, deliberately out of scope:** Tailwind palette
classes map to semantic tokens in a dedicated pass; card/pill component
consolidation; ESLint ratchet. Tracked in
`doc/design/DECISION-SHEET.md`.
## Model Used
Claude Fable 5 (Anthropic, `claude-fable-5`, Mythos-class tier) with
extended thinking, running in Claude Code with tool use; mechanical
phases delegated to Claude Sonnet subagents. Follow-up external-baseline
rework assisted by OpenAI Codex (`gpt-5` coding agent with repository,
terminal, and GitHub tool use). All bulk rewrites executed via
deterministic, idempotent scripts committed in `scripts/`; intentional
visual changes were human-reviewed on screenshot contact sheets.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run targeted local verification and documented the
intentional baseline-publication failure above
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green *(pending new CI run after this
rework)*
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
*(pending review)*
- [x] I will address all Greptile and reviewer comments before
requesting merge
🤖 Generated with [Claude Code](https://claude.com/claude-code) and
OpenAI Codex
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Dotta <bippadotta@protonmail.com>
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent runs often need provider credentials, API tokens, and other
environment-bound secrets.
> - Company-level secrets work for shared credentials, but they do not
model values that should differ by human operator.
> - Without a user-scoped model, a run can dispatch without knowing
whether the responsible human has supplied the needed value.
> - Paperclip also needs run attribution to make those user-scoped
runtime checks deterministic and auditable.
> - This pull request adds user-specific secret definitions, per-user
values, environment bindings, responsible-user attribution, and runtime
resolution gates.
> - The benefit is that teams can define the secret once, let each user
provide their own value, and block runs before dispatch when required
user secrets or active definitions are unavailable.
## Linked Issues or Issue Description
Refs #224
Refs #6057
This PR implements user-specific secret support as a core
secret-management capability rather than a one-off adapter setting. It
is related to existing public work on company secrets UI and runtime
secret refs, but is distinct because the value is owned by the
responsible user and resolved at run dispatch time.
Related PR search before opening found existing secrets work such as
#1550, #8256, #8614, #8634, and #8647; none of those add the full
user-secret definition/value/runtime gate covered here.
## What Changed
- Added user-secret definitions and per-user "My secrets" values,
keeping stored values out of access metadata.
- Added `user_secret_ref` environment bindings and UI affordances to
pick them alongside existing secret refs.
- Added responsible-user runtime resolution so user-secret refs resolve
against the human responsible for the run.
- Added pre-dispatch missing-secret gates so runs fail before adapter
dispatch when required user values are absent or definitions are
inactive.
- Added low-trust allowlist hardening for user-secret runtime access.
- Added issue, routine, run, and agent API key responsible-user
attribution and fail-closed dispatch behavior when attribution cannot be
resolved.
- Added denial-copy mapping so responsible-user authorization failures
surface as actionable run outcomes instead of opaque setup failures.
- Added OpenAPI documentation for the user-secret routes.
- Rebases cleanly on current `master`; migrations were renumbered
incrementally as `0128_user_specific_secrets`,
`0129_agent_api_key_responsible_user`, and
`0130_run_responsible_user_invariant` after upstream `0126`/`0127`
migrations.
- Removed previously committed local design screenshots so the PR
contains code/docs/tests only.
## Verification
- PASS: PR head `2527febd106bcf3ca264ca0da7fca491084192d6` is based on
`paperclipai/paperclip:master`.
- PASS: `git diff --check`
- PASS: `git diff --name-only public/master...HEAD | rg
'^(pnpm-lock\\.yaml|\\.github/workflows/|screenshots/)' || true`
produced no files.
- PASS: migration journal audit confirmed unique indexes through `130`
with tail entries `0126_issue_comment_derived_attribution`,
`0127_environment_custom_images_instance_scoped`,
`0128_user_specific_secrets`, `0129_agent_api_key_responsible_user`, and
`0130_run_responsible_user_invariant`.
- PASS: `pnpm --filter @paperclipai/ui typecheck`
- PASS: `pnpm --filter @paperclipai/server typecheck`
- PASS: `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/heartbeat-responsible-user-invariant.test.ts`
- PASS: `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/heartbeat-active-run-output-watchdog.test.ts
src/__tests__/heartbeat-stale-queue-invalidation.test.ts
src/__tests__/heartbeat-workspace-finalize-branch.test.ts
src/__tests__/issue-monitor-scheduler.test.ts`
- PASS: `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/heartbeat-comment-wake-batching.test.ts
src/__tests__/heartbeat-retry-scheduling.test.ts
src/__tests__/heartbeat-accepted-plan-workspace-refresh.test.ts
src/__tests__/heartbeat-plugin-environment.test.ts`
- PASS: `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/low-trust-red-team-routes.test.ts`
- PASS: `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/secrets-service.test.ts` (55 tests)
- PASS: `pnpm vitest run server/src/__tests__/secrets-routes.test.ts
server/src/__tests__/secrets-service.test.ts` (89 tests after final
Greptile cleanup fixes)
- PASS: `pnpm --filter @paperclipai/server exec vitest run
src/__tests__/heartbeat-issue-liveness-escalation.test.ts` (17 tests
after the final rebase CI fix)
- PASS: focused server Vitest batches covering heartbeat recovery,
project env, plugin env, routines, low-trust, pipelines, monitors,
watchdog, and stale queue paths.
- PASS: GitHub checks are green on
`2527febd106bcf3ca264ca0da7fca491084192d6`, including Typecheck +
Release Registry, Build, General tests, serialized server suites, e2e,
Canary Dry Run, verify, security checks, and Greptile Review.
- PASS: Greptile Review completed successfully on
`2527febd106bcf3ca264ca0da7fca491084192d6` with Confidence Score 5/5,
and GraphQL review-thread audit returned zero unresolved non-outdated
threads.
## Risks
- Runtime behavior now depends on a run having a correct responsible
user; missing or incorrect responsibility assignment can block runs
before adapter dispatch.
- `user_secret_ref` bindings intentionally expose metadata without
values, but UI/API callers may need to handle the new binding kind
explicitly.
- External secret providers and IAM policies are not automatically
provisioned by this PR; operators still need to configure provider-side
access for non-local vaults.
- The PR is broad across db/shared/server/UI/runtime paths, so release
validation should include both API and UI secret workflows before merge.
- The migration renumbering is intentionally incremental after upstream
migrations; the branch migrations use guarded
column/table/index/constraint creation so users who tested the older
draft numbering should not hit duplicate DDL for the existing objects.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex, GPT-5-based coding agent (`gpt-5`), Codex local adapter
with shell/tool use and code execution. Context window and internal
reasoning mode are not exposed by the runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
## Thinking Path
> - Paperclip is the open source control plane people use to manage AI
agents for work.
> - Environment variables and secrets sit in the configuration surfaces
that let agents, projects, routines, and company environments run with
the right runtime inputs.
> - The previous editor was a single legacy component with cramped row
behavior, weak secret conversion affordances, and duplicated handling
across several call sites.
> - Operators need a clearer editor that handles text values, secret
references, draft rows, and sensitive-value warnings consistently
wherever environment variables are configured.
> - This pull request replaces the legacy editor with a reusable
environment variables editor and migrates the existing configuration
surfaces to it.
> - The benefit is a more reliable editing workflow with targeted test
coverage around row state, dotenv parsing, secret selection, and
affected page integrations.
## Linked Issues or Issue Description
No public GitHub issue exists, so this PR describes the issue inline
following the feature request template.
### Subsystem affected
ui/ — React + Vite board UI
### Problem or motivation
Environment variables are edited in several Paperclip configuration
surfaces, including agent config, project properties, stage secrets,
routine sections, company environments, and company settings. The legacy
editor made common operator work difficult: rows could feel cramped,
secret conversion was inconsistent, draft rows and imported dotenv data
were easy to mishandle, and sensitive-value warnings did not have a
consistent place in the workflow.
### Proposed solution
Introduce a reusable environment variables editor component that
consistently supports text values, secret references, draft rows, dotenv
import parsing, sensitive-value hints, secret picking, secret creation,
and conversion to stored secrets. Migrate the existing
environment-variable call sites to the shared editor so behavior and
tests live in one component family.
### Alternatives considered
Keeping the existing `EnvVarEditor` and patching individual call sites
would preserve duplication and leave each surface responsible for its
own row and secret handling. This PR instead centralizes the behavior so
future fixes cover all migrated surfaces.
### Roadmap alignment
Checked `ROADMAP.md`; this does not duplicate a named roadmap item. It
supports the existing local-first and deployment-oriented product
direction by improving the UI where operators configure runtime
environment values.
### Additional context
The PR includes targeted tests for the editor model, dotenv parsing,
sensitive-value detection, component behavior, affected page
integrations, and Greptile review regressions around external saves and
bulk import immutability.
## What Changed
- Replaced the legacy `EnvVarEditor` with a reusable
`environment-variables-editor` component family.
- Added editor model helpers for draft rows, dotenv parsing,
sensitive-value detection, secret picking, secret creation, and
conversion to secret references.
- Migrated agent config, project properties, stage secrets, routine
editable sections, company environments, company settings, design guide
examples, and Storybook stories to the new editor.
- Added targeted tests for the editor model, parsing, sensitive-value
detection, component behavior, and affected company environment/settings
integrations.
- Fixed the company settings test harness to use the repo’s
`flushSync`-based React test helper pattern under the current React
build.
- Addressed Greptile feedback by flushing pending editor drafts before
enclosing form submits or external save-button clicks, cloning
bulk-import rows before mutation, and deferring the overflow
store-as-secret popover open path.
## Verification
- `pnpm exec vitest run
ui/src/components/environment-variables-editor/EnvironmentVariablesEditor.test.tsx
ui/src/pages/CompanyEnvironments.test.tsx
ui/src/components/AgentConfigForm.render.test.tsx` — passed, 3 files /
40 tests.
- Earlier focused Vitest coverage for model, dotenv parsing, sensitive
detection, company environments, and company settings passed, 6 files /
78 tests.
- `pnpm --filter @paperclipai/ui typecheck` — passed.
- GitHub PR checks on head `0aa49c6f8afed1f62d9e26da07fc466fbf299850` —
passed; CI checks green, security review neutral, Greptile check
success.
- Greptile review — 5/5 confidence on head
`0aa49c6f8afed1f62d9e26da07fc466fbf299850`; 0 unresolved review threads.
## Risks
- Medium UI risk: several environment-variable entry surfaces now share
the new editor, so regressions could affect multiple configuration
workflows at once.
- Secret conversion, draft-row behavior, external save flushing, and
bulk import behavior are covered by targeted tests, but reviewer
attention should still focus on manual editing flows, focus retention,
and save/cancel affordances.
- No database or API contract changes.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex, GPT-5-based coding agent with tool use and local command
execution; medium reasoning mode. Exact context-window metadata is not
exposed in this runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Paperclip runs agents through adapters, including built-in local and
gateway-style adapters.
> - Hermes gateway users need to connect Paperclip to an already-running
Hermes API server.
> - The gateway setup flow was missing clear non-local adapter
configuration fields and accepted fewer URL shapes than operators
naturally paste from Hermes.
> - It also surfaced sparse diagnostics when the gateway was unreachable
or when Paperclip generated onboarding prompts for gateway agents.
> - This pull request tightens Hermes gateway configuration, URL
normalization, diagnostics, and onboarding defaults.
> - The benefit is that Hermes gateway setup is easier to complete and
easier to debug without affecting unrelated adapters.
## Linked Issues or Issue Description
No public issue found for this exact follow-up. Related prior/in-flight
Hermes work:
- Refs #2363
- Refs #4359
- Refs #6473
Problem statement:
- **Type:** Adapter follow-up / setup reliability
- **Adapter:** `hermes_gateway`
- **Motivation:** Operators configure `hermes_gateway` against a running
Hermes API server, but the UI and onboarding flow did not expose enough
gateway-specific configuration or diagnostics.
- **Expected behavior:** Paperclip should render the gateway fields,
normalize common Hermes dashboard/API URL inputs, preserve sensible
gateway onboarding defaults, and report reachability failures with
actionable detail.
- **Deployment mode:** Built-in adapter package in the Paperclip
monorepo.
## What Changed
- Added UI config fields for non-local Hermes gateway settings,
including tests for rendering and field behavior.
- Accepted Hermes dashboard URLs by normalizing them to gateway API URLs
for execution.
- Improved gateway reachability and run URL diagnostics.
- Updated Hermes gateway onboarding text/default behavior so join
prompts preserve gateway configuration.
- Added focused server, UI, and adapter tests for the gateway
configuration and onboarding paths.
## Verification
- `pnpm install --frozen-lockfile --prefer-offline`
- `pnpm --filter @paperclipai/hermes-paperclip-adapter exec vitest run
src/gateway/server/execute.test.ts` — 20 passed
- `pnpm exec vitest run
server/src/__tests__/invite-accept-gateway-defaults.test.ts
server/src/__tests__/invite-onboarding-text.test.ts
ui/src/adapters/hermes-gateway/config-fields.test.tsx
ui/src/components/AgentConfigForm.render.test.tsx
ui/src/lib/agent-onboarding-prompt.test.ts` — 23 passed across 5 files
- Confirmed the PR diff excludes `pnpm-lock.yaml` and
`.github/workflows`.
## Risks
Low to moderate risk. The changes are scoped to Hermes gateway
configuration/onboarding and generic non-local adapter field rendering.
The main risk is rejecting an unusual Hermes URL shape that should be
accepted; the normalization tests cover dashboard and API URL variants
added here.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex, GPT-5-based coding agent in a tool-enabled local CLI
environment, with shell/GitHub/Paperclip API access. Exact runtime model
identifier was not exposed in this session.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent adapters are the boundary between the control plane and the
runtimes that actually do work.
> - Hermes support needs to be available as first-class local and
gateway adapters while still preserving the adapter-manager override
path for external packages.
> - The adapter work touches runtime execution, UI adapter metadata,
onboarding prompts, scoped credentials, release packaging, and smoke
coverage, so the handoff needs concrete verification rather than only
unit tests.
> - This pull request adds built-in Hermes local and Hermes gateway
support, keeps external adapter overrides compatible, and
documents/tests the gateway flow end to end.
> - The benefit is that operators can hire Hermes-backed agents without
a manual plugin install, while self-hosted installs can still
override/shadow the built-ins through Adapter manager packages.
## Linked Issues or Issue Description
No public GitHub issue exists for this exact Hermes built-in adapter,
gateway onboarding, and release-source work.
Problem description:
- Hermes local and gateway adapters need a public, reviewable source
path in the monorepo so package artifacts and built-in adapter behavior
match the application source.
- Operators need built-in `hermes_local` and `hermes_gateway` adapter
choices without losing the ability to install external Hermes packages
as overrides.
- Gateway onboarding needs secure defaults for API server URLs, API
keys, and generated agent setup text.
- Hermes-originated task bridge credentials need narrower API-key scope
configuration.
- Related public PRs found during duplicate search include #3027, #2363,
#7544, #7950, #8095, and #8543.
## What Changed
- Added the unified Hermes adapter package with local and gateway
server/UI/CLI exports, config schemas, transcript parsing, model
detection, and package metadata.
- Registered `hermes_local` and `hermes_gateway` as built-in adapters
across shared constants, server registries, CLI packaging, and UI
adapter registries.
- Kept the external adapter override path compatible so installed Hermes
packages can shadow built-ins and restore the built-in parser when
disabled.
- Added Hermes gateway onboarding docs, board-operator docs, Docker
smoke assets, and shell smoke harnesses for join/e2e validation.
- Added scoped task-bridge API-key support, authorization checks,
issue-origin handling, and tests for Hermes-created Paperclip tasks.
- Hardened gateway transport and redaction behavior for API keys,
headers, session data, and smoke diagnostics.
- Updated release packaging/bootstrap checks for the Hermes packages
while leaving `pnpm-lock.yaml` out of the PR per repository policy.
## Verification
Targeted local verification recorded before PR handoff:
- `pnpm --filter @paperclipai/hermes-paperclip-adapter exec vitest run
src/gateway/server/execute.test.ts` — 14/14 passed.
- `pnpm test:hermes-gateway-smoke` — 6/6 passed.
- Hermes package typecheck/build checks passed.
- Focused server/UI adapter tests passed — 31/31.
- Release helper Node tests passed — 18/18.
- `git diff --check origin/master..HEAD` passed.
Fresh Docker E2E smoke evidence:
- Ran `pnpm smoke:hermes-gateway-e2e` on 2026-06-26 with a fresh state
directory and fresh Docker container against a live Paperclip dev
server.
- Hermes direct execution reached `completed`.
- Hermes stop/cancel path reached `cancelled`.
- Hermes gateway created a Paperclip task, Paperclip ran the Hermes
agent, and the task reached `done` with the expected marker response.
- Temporary board auth keys, token files, smoke state, and Docker
containers were cleaned up after the run.
PR checks on head `b5eae40ce`:
- GitHub Actions passed: `policy`, `review`, `Typecheck + Release
Registry`, all general test shards, all serialized server shards,
`Build`, `Canary Dry Run`, `e2e`, and aggregate `verify`.
- External checks passed: Snyk and Socket Project Report.
- External Socket Pull Request Alerts remained pending after the
first-party CI matrix completed.
## Risks
- Medium risk: this spans adapter registration, package publishing,
gateway execution, onboarding docs, API-key scoping, and UI adapter
metadata.
- Migration risk is low: the scope-config migration adds a nullable
column and does not rewrite existing keys.
- Gateway execution depends on operator-provided Hermes API
configuration; the smoke covers the Docker gateway path but real
deployments may differ by network/auth setup.
- Direct Greptile review on the latest expanded diff is file-count
limited, although the commitperclip review gate passed.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
OpenAI Codex, GPT-5 coding agent, tool use enabled in a local repository
workspace. Context window size is not exposed in this environment.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Commitperclip review gate is green; direct Greptile review is
file-count limited on the latest expanded diff
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent configuration includes adapter model settings and optional
cheap model profiles used by runtime lanes.
> - The agent configuration screen already exposes a Test action for
adapter environment checks.
> - When a cheap model profile is configured, that Test action only
exercised the primary model configuration.
> - That left users able to save a cheap model profile that had not been
validated by the same configuration test flow.
> - This pull request makes the Test action probe both the primary model
and the configured cheap model.
> - The benefit is earlier feedback when the cheap model is unavailable
or misconfigured.
## Linked Issues or Issue Description
No exact public issue found.
Problem description:
- What happened: the agent configuration Test action validated the
primary adapter model but did not validate an enabled cheap model
profile.
- Expected behavior: when a cheap model is configured and enabled, the
Test action should also test that cheap model using the same environment
selection.
- Steps to reproduce: configure an agent with a primary model and
enabled cheap model profile, then click Test in the agent configuration
UI.
- Paperclip version/commit: current `master` at PR creation.
- Deployment mode: local development UI behavior.
Related public context found during GitHub search:
- #4881 added cheap model profiles for local adapters.
- #6534 is related cheap-primary-model preservation work.
- #6956 tracks broader agent configuration settings exposure.
GitHub searches performed:
- `cheap model config test`
- `AgentConfigForm cheap model`
- `modelProfiles cheap`
- `adapter environment cheap model`
## What Changed
- Updated `AgentConfigForm` so the adapter environment Test action runs
the primary model check first.
- Added a second cheap model check when the cheap profile is enabled and
resolves to a model.
- Built the cheap test payload from the resolved cheap profile config
instead of a model-only override, preserving adapter-default and saved
cheap-profile fields.
- Merged the individual results into a single labeled result so the UI
can show both outcomes together.
- Preserved request/API failures so they still surface through the
existing error UI instead of becoming synthetic adapter checks.
- Added render coverage asserting both primary and cheap test calls,
non-model cheap-profile fields, and request failure handling.
## Verification
- `git diff origin/master...HEAD | rg -n
"(API[_-]?KEY|SECRET|TOKEN|PASSWORD|PRIVATE[_-]?KEY|BEGIN RSA|BEGIN
OPENSSH|Bearer
[A-Za-z0-9._-]+|ghp_[A-Za-z0-9_]+|sk-[A-Za-z0-9]+|AIza[0-9A-Za-z_-]+|OPENAI_API_KEY|ANTHROPIC_API_KEY)"
|| true` produced no matches.
- `git diff --check origin/master...HEAD`
- `pnpm --filter @paperclipai/ui exec vitest run
src/components/AgentConfigForm.render.test.tsx --reporter=dot`
- `pnpm --filter @paperclipai/ui typecheck`
- GitHub PR checks on head `3d9ba15d2` are green, including Build,
Typecheck + Release Registry, General tests, serialized server suites,
e2e, Canary Dry Run, security scans, and policy/review checks.
- Greptile Review passed on head `3d9ba15d2`; all review threads are
resolved.
## Risks
Low risk. The change is limited to the agent configuration UI test
action and its render test. The cheap-model probe now preserves
adapter-default and saved cheap-profile fields, matching the runtime
merge order more closely. The main residual risk is adapter-specific UI
coverage for cheap-profile fields that are stored but not directly
editable in this form.
## Model Used
OpenAI GPT-5 Codex via the Codex local agent environment, with
terminal/tool use for repository inspection, implementation, GitHub CLI
operations, and local verification. Exact context window was not exposed
by the runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent configuration is where operators set identity, adapter
behavior, and execution environment defaults.
> - The environment override control should only appear when there is a
meaningful choice beyond the always-available local environment, or when
an existing saved override must remain visible so it can be inspected or
cleared.
> - The previous UI used an "Execution" header, explanatory helper copy,
and verbose inherited-default wording that made the agent config feel
noisier than necessary.
> - This pull request narrows the selector visibility to real
alternatives and forced Kubernetes mode, then updates the visible copy
to match the environment-focused mental model.
> - The benefit is a cleaner agent configuration surface that only asks
operators to make an environment choice when that choice exists.
## Linked Issues or Issue Description
No public GitHub issue exists for this change. Public GitHub searches
for related issues and PRs returned no matches:
- `agent config environment selector`
- `execution environment agent config`
Bug-style issue description:
What happened?
The agent config form could surface an environment override section even
when the operator had no meaningful non-local execution environment
choice. The section was labeled "Execution", included helper/inheritance
text, and described the inherited local default as "Inherit instance
default (Local)".
Expected behavior:
The selector should stay hidden unless there is more than one configured
environment choice, the section should be labeled "Environment", helper
text should be removed, and the inherited default option should read
like `Default: Local`. Existing saved non-local overrides should remain
visible even if the target environment is no longer runnable, so
operators can inspect or clear the stale selection.
Steps to reproduce:
1. Run Paperclip from `master` with environments enabled and only the
implicit local default, or Local plus one runnable non-local
environment.
2. Open an agent configuration form.
3. Inspect the environment override section visibility and copy.
Paperclip version or commit: `0b945f449` / current `master` before this
branch.
Deployment mode: Local dev (`pnpm dev`).
Installation method: Built from source.
Agent adapters involved: Not adapter-specific.
Database mode: Not database-related.
Access context: Board operator UI.
## What Changed
- Shows the environment override selector only for forced Kubernetes
mode, at least one runnable non-local environment, or an existing saved
non-local override.
- Keeps Local out of the runnable environment count so Local-only setups
do not show a redundant selector.
- Preserves stale saved non-local overrides in the selector options so
operators can see and clear them.
- Renames the section header from "Execution" to "Environment".
- Removes the helper/inheritance text above the selector.
- Changes the inherited option copy from `Inherit instance default
(...)` to `Default: ...`.
- Adds focused render tests for Local-only hiding, Local plus one
runnable non-local environment showing the concise selector copy, and
stale non-runnable override recovery.
## Verification
- `pnpm --filter @paperclipai/ui exec vitest run
src/components/AgentConfigForm.render.test.tsx`
- `pnpm --filter @paperclipai/ui typecheck`
- `git diff --check origin/master...HEAD`
- Checked `ROADMAP.md` for overlapping planned core work.
- Searched public GitHub issues and PRs for duplicate or related work;
no matches found.
## Screenshots
Before, when the selector was visible:
```text
Execution
Environment override
Inheriting the instance default: Local.
[Inherit instance default (Local)]
```
After, when Local plus a runnable non-local environment exists:
```text
Environment
Environment override
[Default: Local]
[E2B · sandbox]
```
After, when only Local is configured:
```text
(no Environment section is rendered)
```
## Risks
Low risk. This is isolated to the agent config UI. The main behavioral
risk is hiding the selector in an environment edge case; the current
branch keeps forced Kubernetes mode, runnable non-local environments,
and pre-existing saved non-local overrides visible. The added render
tests cover the Local-only, single non-local alternative, and stale
override cases.
## Model Used
OpenAI Codex local adapter, GPT-5-based Codex coding session. The exact
hosted backend model ID is not exposed in the runtime. Tool use and
local shell execution were enabled.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agent adapter configuration is where operators edit the runtime
settings that control how each agent is invoked.
> - The adapter form includes a test action so operators can validate
the in-progress configuration before saving it.
> - In edit mode, that test action changed into `Save + Test` when the
form was dirty, which mixed validation with persistence.
> - Operators already have an explicit Save action for committing
adapter changes.
> - This pull request makes the adapter test action test-only so
validation never implicitly saves a dirty draft.
> - The benefit is clearer operator control: Test validates the current
draft, Save persists it.
## Linked Issues or Issue Description
No matching public GitHub issue or in-flight PR was found, so this PR
describes the bug inline.
### Pre-submission checklist
- I searched existing open and closed issues and did not find a
duplicate.
- I can reproduce this behavior on current `master` before this branch.
- I confirmed this originates in the Paperclip UI, not in a specific
local agent CLI or provider configuration.
### What happened?
When editing an existing agent adapter configuration, changing an
adapter setting made the adapter environment test button change from
`Test` to `Save + Test`. Clicking it saved dirty edits before running
the environment test.
### Expected behavior
The test action should always be a test action. Dirty draft values
should be validated without persisting them, and the explicit Save
button should be the only save path.
### Steps to reproduce
1. Open an existing agent adapter configuration.
2. Edit an adapter setting so the form becomes dirty.
3. Observe the adapter environment test action label and behavior.
### Paperclip version or commit
Current `master` before this branch, `c0743482b`.
### Deployment mode
Local dev/private operator UI.
### Installation method
Built from source with pnpm.
### Agent adapter(s) involved
Not adapter-specific. The behavior lives in the shared agent adapter
config form.
### Database mode
Not database-related.
### Access context
Board operator UI.
### Additional context
Public searches performed before opening this PR:
- `adapter test save`
- `Save + Test`
- `AgentConfigForm testEnvironment`
- `adapter configuration test button`
No logs or local config are needed for this UI-only behavior report.
## What Changed
- Removed the helper that converted dirty edit-mode adapter tests into
`Save + Test`.
- Removed the save-before-test helper path so clicking Test only invokes
the adapter environment test mutation.
- Kept draft testing behavior by continuing to build the test payload
from the current in-progress adapter config.
- Dropped unit tests that asserted the old save-and-test behavior.
## Verification
- `pnpm vitest run ui/src/components/AgentConfigForm.test.ts` -> 6/6
passed.
- `pnpm --filter @paperclipai/ui typecheck` was attempted locally and
failed on the existing unrelated
`packages/plugins/sdk/src/ui/components.ts` React module-resolution
error before reaching this change.
- Static review: `Save + Test`, `getAgentConfigTestActionLabel`, and
`runAgentConfigEnvironmentTest` no longer appear in `ui/src`.
- Before/after visual state for the dirty edit-mode adapter test action:
- Before: `Save + Test`
- After: `Test`
- PR CI passed: policy, commitperclip review,
typecheck/release-registry, build, general tests, serialized server
suites, e2e, canary dry run, aggregate verify, and security scans.
- Greptile passed with 5/5 confidence. The remaining
screenshot/documentation thread was answered and resolved because this
is a text-only button-label state already captured above.
## Risks
Low risk. This is a narrow UI behavior change in the adapter config
form. The main behavior change is intentional: testing a dirty adapter
draft no longer persists it. If an operator expected Test to save edits,
they must now use the explicit Save button.
## Model Used
- Anthropic Claude via Paperclip `claude_local` produced the initial
implementation commit.
- OpenAI GPT-5 Codex via Paperclip `codex_local` reviewed the change,
amended public commit metadata, pushed the branch, and prepared this
pull request. The adapter did not expose an exact context-window value
in the run payload; tool use and local shell execution were used.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] If this change affects the UI, I have included before/after
screenshots
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
Co-authored-by: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the control plane for AI-agent companies, so execution
environment selection has to stay inspectable and predictable across
companies, agents, and runs.
> - The environment subsystem decides where an agent heartbeat actually
runs and how remote sandbox state is realized and restored.
> - That subsystem previously mixed company-scoped environment catalogs
with issue-level environment stamping, so a reassigned issue could keep
executing in the previous assignee's sandbox.
> - That behavior breaks the control-plane contract: changing the
assignee should change the executing agent/environment path unless there
is an explicit current override.
> - Fixing it cleanly required more than a narrow patch; the environment
model had to move to instance scope with a single inherited default and
per-agent override semantics.
> - This pull request rewires the schema, server/API surface, runtime
resolution, and UI around that model, then adds regression coverage for
cross-company inheritance and per-agent isolation.
> - The benefit is that environment choice now follows the approved
instance/agent configuration path instead of stale issue state, while
shared environments only need to be configured once per instance.
## Linked Issues or Issue Description
- No directly matching public GitHub issue or PR was found while
searching for this refactor.
### What happened?
Reassigning work between agents with different execution environments
could keep running in the previous sandbox because environment choice
was stamped onto the issue and outranked the current assignee. The same
subsystem also forced environment catalogs to be duplicated per company
even though the underlying execution environments were instance-wide
resources.
### Expected behavior
Execution should resolve through the current instance and agent
configuration path, with one instance-scoped environment catalog, one
instance default, optional per-agent override, and no stale issue-level
environment authority surviving reassignment.
### Steps to reproduce
1. Configure two agents to use different execution environments.
2. Assign an issue to the first agent so the issue records execution
state in that environment.
3. Reassign the same issue to the second agent and run another
heartbeat.
4. Observe that the pre-fix runtime can still sync or execute in the
original sandbox instead of the second agent's environment.
### Paperclip version or commit
Current `master` before this PR.
### Deployment mode
Self-hosted server.
### Installation method
Built from source (`pnpm dev` / `pnpm build`).
### Agent adapter(s) involved
- Claude Code
- Not adapter-specific (core bug in environment authority / resolution)
### Database mode
External Postgres.
### Access context
Both board reassignment and agent heartbeats were involved.
## What Changed
- Moved environments and their default selection contract to instance
scope in DB/shared types, including the migration that dedupes legacy
per-company environments and seeds the instance local default.
- Reworked environment CRUD/auth flows to use instance-scoped APIs and
added route/service coverage for instance-level environment management.
- Changed runtime resolution to prefer `agent default -> instance
default -> built-in local`, removed issue-level environment stamping
from the active execution path, and isolated sandbox/plugin leases by
`(executionWorkspaceId, agentId)`.
- Added environment env-var runtime precedence so environment-provided
values act as the baseline for agent execution.
- Moved the environment UI into instance settings and updated agent
configuration surfaces to reflect inherit/override behavior.
- Added regression coverage for instance-default inheritance across
companies and for the new runtime resolution behavior.
- Fixed a rebase-only duplicate `enableTaskWatchdogs` flag regression in
instance settings types/validators/services so the branch typechecks
cleanly on current `master`.
- Updated stale server tests so CI matches the shipped instance-scoped
environment contract.
## Verification
- `git diff --check`
- `pnpm --filter @paperclipai/shared typecheck`
- `pnpm --filter @paperclipai/db typecheck`
- `pnpm exec vitest run
server/src/__tests__/environment-runtime-driver-contract.test.ts
server/src/__tests__/agent-permissions-routes.test.ts
server/src/__tests__/environment-routes.test.ts
server/src/__tests__/environment-instance-routes.test.ts
server/src/__tests__/execution-workspace-policy.test.ts
server/src/__tests__/heartbeat-plugin-environment.test.ts
server/src/__tests__/instance-settings-routes.test.ts`
## Risks
- The migration changes environment scope and dedupes existing rows, so
installs with unusual legacy environment combinations should be reviewed
carefully during upgrade.
- Remote execution behavior now depends on instance-default inheritance
semantics instead of issue-level stamping, so any remaining code paths
that still assume issue-scoped environment authority would surface as
follow-up bugs.
- This PR includes both server/runtime behavior and UI relocation, so
reviewers should watch for authorization edge cases around instance
settings and environment management.
> I checked [`ROADMAP.md`](ROADMAP.md). This work fits the existing
Cloud / Sandbox agents direction as a bug-fix/refactor to current
behavior, not a new parallel product surface.
## Model Used
- OpenAI Codex coding agent in this Paperclip/Codex session; GPT-5-class
tool-using model with code execution and shell access. The exact backend
model ID is not exposed to the session runtime.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [ ] If this change affects the UI, I have included before/after
screenshots
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Creating an agent starts at **New Agent → "manually" → pick an
adapter**, which routes to
`/{company}/agents/new?adapterType=claude_local` and renders the
`NewAgent` page with the `AgentConfigForm`
> - `AgentConfigForm` hands the parent a `triggerTestEnvironment`
callback via an `onTestActionChange` effect so the page can wire up its
"Test"/"Save + Test" button
> - That trigger was rebuilt on every render: it depended on
`runEnvironmentTest`, which is derived from a react-query `useMutation`
result, and `useMutation` returns a **brand-new result object identity
on every render**
> - So the `onTestActionChange` effect re-fired every render and pushed
a new function into the parent's state, producing an infinite `setState`
loop ("Maximum update depth exceeded") that threw during render
> - The app's custom router updates location **without remounting**, and
there was no error boundary around the routed outlet, so the throw left
a dead render tree — a fully **blank page** that stayed blank on
back-navigation until a hard refresh
> - This pull request stabilizes the trigger with a latest-ref pattern
so the effect no longer re-fires, and adds a route-keyed error boundary
so any future render throw degrades to a recoverable error card instead
of a blank screen
> - The benefit is that creating an agent works again, and render-time
failures anywhere in the routed UI are contained and recoverable rather
than silently blanking the app
## Linked Issues or Issue Description
No public GitHub issue exists, so the underlying bug is described inline
following the bug report template
(`.github/ISSUE_TEMPLATE/bug_report.yml`):
### What happened?
In the UI, choosing **New Agent → "manually" → (any adapter, e.g.
Claude)** navigates to the agent-config page and renders a **completely
blank page**. The browser console shows React's `Maximum update depth
exceeded`. Using the back button changes the URL but the page stays
blank until a full hard refresh.
### Expected behavior
Selecting an adapter shows the agent configuration form so the agent can
be created.
### Steps to reproduce
1. Open the app and click **New Agent**.
2. Choose **manually**.
3. Pick an adapter (e.g. Claude / `claude_local`).
4. Observe the blank page (URL becomes
`/{company}/agents/new?adapterType=claude_local`).
### Paperclip version or commit
Reproduces on `master` (base of this PR).
### Deployment mode
Reproduces regardless of deployment mode — it is a client-side render
loop.
### Root cause
`useMutation` returns a new result object identity each render, so the
`runEnvironmentTest`-derived `triggerTestEnvironment` callback was
unstable, which made the `onTestActionChange` effect push a new function
into parent state every render → infinite update loop → render throw →
no boundary → blank tree.
## What Changed
- **`ui/src/components/AgentConfigForm.tsx`** — Stabilize the
environment-test trigger handed to the parent using a latest-ref
pattern: the churny behavior (`runEnvironmentTest`,
`testEnvironmentDisabled`) lives in a `useRef` updated by an effect, and
the exposed `triggerTestEnvironment` is a `useCallback(() =>
triggerRef.current(), [])` with an empty dep array, so its identity is
stable across renders and the `onTestActionChange` effect no longer
re-fires every render.
- **`ui/src/components/RouteErrorBoundary.tsx`** (new) — A route-keyed
React error boundary that catches render throws and renders a
recoverable error card (showing the error message, with "Go back" and
"Reload page" actions). It resets automatically when the route
(`pathname + search`) changes.
- **`ui/src/components/Layout.tsx`** — Wrap the routed `<Outlet />` in
`<RouteErrorBoundary>` so a render throw degrades to the error card
instead of a blank page.
- **`ui/src/components/RouteErrorBoundary.test.tsx`** (new) — Regression
test: a throwing child is contained as a recoverable error card (showing
the message), "Go back" calls `navigate(-1)`, and the boundary resets to
render children again after the route changes.
## Verification
- `npx vitest run ui/src/components/RouteErrorBoundary.test.tsx` → 3
passed (regression test for this fix).
- `npx vitest run ui/src/components/AgentConfigForm.test.ts` → 9 passed.
- `npx tsc -b` in `ui` → 0 errors.
- Manual: ran the dev server, clicked **New Agent → manually → Claude**.
- **Before:** blank page; console logs `Maximum update depth exceeded`;
back button leaves the page blank until hard refresh.
- **After:** the agent configuration form renders normally and the agent
can be created; navigating away and back works without a hard refresh.
- Boundary check: with the loop still in place (pre-fix), the new
boundary catches the throw and shows a recoverable error card instead of
a blank screen; "Go back" / route change resets it.
_Screenshots: the before-state is the React `Maximum update depth
exceeded` error and a blank `/agents/new` page; the after-state is the
rendered agent-config form. Both were observed locally; rendered images
can be attached on request._
## Risks
- **Low risk.** Changes are confined to three UI files with no API,
schema, or behavioral change to agent creation beyond fixing the loop.
- The latest-ref pattern preserves identical runtime behavior of the
test trigger (same guard, same `runEnvironmentTest()` call) — it only
stabilizes the callback identity.
- The error boundary is additive; on the happy path it renders its
children unchanged. Its only behavior is to catch render throws that
previously blanked the app.
## Model Used
Claude (Anthropic), model `claude-opus-4-8` — extended-thinking-capable,
tool-use (file edit, shell, tests). Used to diagnose the infinite render
loop, implement the latest-ref fix and route error boundary, and verify
via local typecheck/tests.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [ ] If this change affects the UI, I have included before/after
screenshots (described textually in Verification — see note)
- [ ] I have updated relevant documentation to reflect my changes (N/A —
bug fix, no docs affected)
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
## Thinking Path
> - Paperclip is the control plane people use to create, configure, and
run AI agents for work.
> - This change sits in the server-side agent lifecycle and
secret-binding subsystem, where adapter config `env` entries can
reference company secrets.
> - An incident (while trying to configure a Novita sandbox) showed that
an agent can reach a broken runtime state if `adapterConfig.env`
contains `secret_ref` entries but the matching `company_secret_bindings`
rows are missing.
> - The immediate run-path guard and error-surfacing work made the
failure diagnosable, but they did not fully prevent new broken agents
from being created.
> - The risk came from create and approval flows being responsible for
remembering to sync bindings at each call site, which is easy to miss as
new flows are added.
> - This pull request moves the invariant into `agentService`
create/update/activate paths, keeps the existing hire-flow fix, and adds
regression coverage for create, update, and legacy pending-approval
recovery.
> - The benefit is that agent secret binding integrity is enforced
closer to the data mutation point, so future callers inherit the
protection automatically.
## Linked Issues or Issue Description
Refs #8309
### What happened?
A Paperclip agent could persist `adapterConfig.env` `secret_ref` entries
without matching agent-scoped `company_secret_bindings` rows. When that
happened, the config UI could still look configured, but the real run
path failed pre-dispatch because the secret was not actually bound to
that agent.
### Expected behavior
Every normal agent create, config-update, and pending-approval
activation flow should leave the agent with secret bindings that match
its persisted secret-ref env config.
### Steps to reproduce
1. Create or activate an agent through a flow that persists
`adapterConfig.env` secret refs without synchronizing
`company_secret_bindings`.
2. Observe that the config state can still appear populated.
3. Start a run for that agent.
4. Observe that pre-dispatch binding validation fails because the secret
reference exists but the agent binding does not.
### Deployment mode
Local dev (`pnpm dev`)
### Installation method
Built from source (`pnpm dev` / `pnpm build`)
### Agent adapter(s) involved
- Claude Code
- Not adapter-specific (core bug)
### Database mode
Embedded PGlite / embedded local dev database flow
### Access context
Board (human operator) created or approved the agent; agent runtime
later consumed the config.
### Additional context
This PR focuses on preventing new broken states from normal service
flows and on backfilling the covered legacy pending-approval activation
path.
## What Changed
- Kept the existing branch-local hire-flow fix that synchronized
bindings for route and approval paths.
- Moved the binding integrity invariant into `agentService.create()`,
`agentService.update()` when `adapterConfig` changes, and
`agentService.activatePendingApproval()`.
- Added `server/src/__tests__/agents-service-secret-bindings.test.ts`
covering create-time sync, update-time resync, and backfill for legacy
pending-approval agents.
- Removed now-redundant route-layer and approval-layer binding sync
calls once the service layer became authoritative.
- Simplified the affected unit tests so route/approval tests no longer
assert service-owned binding writes directly.
## Verification
- `pnpm --filter @paperclipai/server typecheck`
- `pnpm exec vitest run
server/src/__tests__/agents-service-secret-bindings.test.ts
server/src/__tests__/approvals-service.test.ts
server/src/__tests__/agent-skills-routes.test.ts`
## Risks
- Low to medium risk.
- This changes where secret-binding synchronization is enforced, so any
unexpected caller that relied on upper-layer manual sync behavior could
behave differently.
- Agent create/update/activation flows now perform binding
synchronization consistently, which adds binding-table writes at those
mutation points.
- This PR does not retroactively scan and heal every already-broken
historical agent row; it prevents and backfills through the covered
service flows.
> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.
## Model Used
- OpenAI Codex / GPT-5 Codex class model via `codex_local`
- Session model family: GPT-5 Codex
- Tool-assisted coding with shell, git, HTTP, and local test execution
- Reasoning mode: medium interactive tool-use workflow
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] If this change affects the UI, I have included before/after
screenshots
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Nicky Leach
Tonio
Devin Foley
Dotta
Michael Nguyen
scotttong
David V
