+
Agentic OS โ Built with โค๏ธ by Mihir Modi
+ ๐ Read the dev.to article
+ ยท
Report Bug
ยท
Discussions
diff --git a/docs/devto-article.md b/docs/devto-article.md
index 2322c1b..6a1ad7e 100644
--- a/docs/devto-article.md
+++ b/docs/devto-article.md
@@ -1,68 +1,48 @@
---
-title: I Built an Open-Source Agent OS โ Here's the Architecture
-published: false
-description: "How I built Agentic OS: a locally-hosted multi-agent orchestration platform that coordinates opencode, Hermes Agent, and Gemini CLI into one dashboard with 16+ skills, cron scheduler, cost analytics, and persistent memory. Full architecture breakdown."
+title: I Built an Open-Source Agent OS
+published: true
+description: "A locally-hosted multi-agent OS that orchestrates opencode, Hermes Agent, and Gemini CLI into one dashboard with 16 skills, scheduler, cost analytics, and memory. Full architecture breakdown."
tags: [agents, devops, opensource, python, fastapi]
cover_image: https://raw.githubusercontent.com/modimihir07/agentic-os/main/docs/og-image.png
+canonical_url: https://dev.to/mihir_nmodi_14a06a4019e1/i-built-an-open-source-agent-os-2h30
---
-# I Built an Open-Source Agent OS โ Here's the Architecture
+I built an operating system for AI agents.
-Most AI agent tools work in isolation. You have a terminal for coding, a separate chat for research, another dashboard for monitoring, and yet another tool for memory. Each one is powerful alone, but together? Chaos.
+Not a terminal. Not a chat app. A full dashboard that coordinates three agents โ **opencode** (code/DevOps), **Hermes Agent** (memory/scheduling), and **Gemini CLI** (research) โ into one unified control plane.
-That's why I built **Agentic OS** โ a locally-hosted operating system for AI agents that coordinates three agents (opencode, Hermes Agent, and Gemini CLI) into one unified dashboard.
+GitHub: [github.com/modimihir07/agentic-os](https://github.com/modimihir07/agentic-os)
-
-
-**GitHub**: [github.com/modimihir07/agentic-os](https://github.com/modimihir07/agentic-os)
+---
## The Problem
-I was using:
-- **opencode** for code generation and DevOps tasks
-- **Hermes Agent** for persistent memory and scheduling
-- **Gemini CLI** for web research and analysis
+I was jumping between three tools:
+- opencode for coding tasks
+- Hermes for scheduling and memory
+- Gemini CLI for web research
-Each was great at its job, but there was no central hub. No shared memory. No unified dashboard. No way to chain them together. Every session started from zero.
+No shared context. No central dashboard. Every session started from zero.
## The Solution: 7-Layer Architecture
-Instead of a flat "three agents" approach, I designed a **7-layer architecture** that standardizes how agents interact:
-
```
-Layer 7: Identity / Persona / Constitution
-Layer 6: Self-Evolution + Capability Manager
-Layer 5: Scheduler + Awareness + Health Guardian
-Layer 4: Memory Graph + Memory Consolidation
-Layer 3: Skills Hub + Eval + Learnings Loop
-Layer 2: Business Brain + Context Folders
-Layer 1: Agent Router + Standards + Profiles
+Layer 7 Identity / Constitution
+Layer 6 Self-Evolution
+Layer 5 Scheduler + Health
+Layer 4 Memory Graph
+Layer 3 Skills Hub + Eval
+Layer 2 Business Brain
+Layer 1 Agent Router
```
-### Layer 1: Agent Router
-Routes tasks to the right agent automatically:
-- **Code/DevOps task?** โ opencode
-- **Memory/Scheduling?** โ Hermes
-- **Research/Analysis?** โ Gemini CLI
-- **Complex multi-step?** โ Chain: Gemini researches โ opencode implements โ Hermes monitors
+**Layer 1** routes tasks automatically: code โ opencode, memory โ Hermes, research โ Gemini.
-### Layer 3: Skills Hub (16 Skills)
-Each skill has a standardized structure:
-```
-skills/skill-name/
-โโโ SKILL.md # Instructions
-โโโ learnings.md # Accumulated knowledge
-โโโ eval.json # Scoring criteria
-โโโ score-history.json # Performance tracking
-โโโ context/ # Task inputs
-```
+**Layer 3** gives each skill a standardized structure (SKILL.md + learnings + eval + score history) so they improve over time.
-This means skills improve over time โ the more you run them, the smarter they get.
+**Layer 4** provides persistent memory via a shared `brain/` folder and SQLite FTS5 โ agents remember context across sessions.
-### Layer 4: Persistent Memory
-All agents read from a shared `brain/` folder at session start. Hermes provides SQLite FTS5 for full-text search across all past conversations.
-
-## Feature Highlights
+## 12 Features
| Feature | What it does |
|---------|-------------|
@@ -74,51 +54,38 @@ All agents read from a shared `brain/` folder at session start. Hermes provides
| Audit Trail | Every action logged |
| Prompt Library | 10 reusable templates |
| Dark/Light Theme | GitHub-style dark mode |
+| Standards System | Auto-discover project conventions |
+| Plugin Registry | Extend via custom skills |
+| Client Timeout | 200s AbortController for long queries |
+| Memory | SQLite FTS5 + shared brain/ files |
-## Comparison with Julian Goldie's Claude Agent OS
+## vs Claude Agent OS
-I was inspired by Julian Goldie's "Agent OS: Claude + Hermes = Superpowers" video. Here's an honest comparison:
+Honest comparison with Julian Goldie's video project:
| Feature | Claude Agent OS | Agentic OS (Mine) |
|---------|----------------|-------------------|
-| Core Agents | Claude + OpenClaw + Hermes | opencode + Hermes + Gemini |
-| Cost | $20/mo (Claude) | **$0 (all free tiers)** |
-| Stack | Next.js + Tailwind | FastAPI + vanilla JS |
-| Memory | Obsidian (external) | **Built-in brain/ + SQLite** |
-| Cost Tracking | None | **Built-in analytics** |
-| Backup/Restore | None | **One-click** |
-| Open Source | No | **MIT License** |
+| Agents | Claude + OpenClaw + Hermes | opencode + Hermes + Gemini |
+| Cost | $20/mo | **$0 (all free tiers)** |
+| Memory | Obsidian (external) | **Built-in** |
+| Cost Tracking | None | **Built-in** |
+| Backup | None | **One-click** |
+| License | Tutorial only | **MIT** |
## Tech Stack
-- **Backend**: FastAPI (Python)
-- **Frontend**: Vanilla JS SPA (no framework)
-- **Scheduler**: APScheduler
-- **Memory**: SQLite FTS5 + Markdown files
-- **Agents**: opencode, Hermes Agent, Gemini CLI
-- **Cost**: $0 โ all free tiers (Gemini Flash, OpenRouter free models)
+Backend: FastAPI (Python) ยท Frontend: vanilla JS SPA ยท Scheduler: APScheduler ยท Memory: SQLite FTS5 ยท Cost: $0 (Gemini Flash, OpenRouter free models)
## Quick Start
```bash
git clone https://github.com/modimihir07/agentic-os.git
-cd agentic-os
-chmod +x install.sh && ./install.sh
-./start.sh
+cd agentic-os && ./install.sh && ./start.sh
# Open http://127.0.0.1:8080
```
-## What's Next
-
-I'm planning to add:
-- More skills (Kubernetes monitoring, CI/CD pipelines)
-- Agent handoff protocol for complex multi-step tasks
-- Web search integration across all agents
-
---
-**Agentic OS** is open-source under MIT. Star it on GitHub if you find it useful!
+Agentic OS is MIT licensed. Star it on GitHub if you find it useful.
๐ **[github.com/modimihir07/agentic-os](https://github.com/modimihir07/agentic-os)**
-
-Questions? Comments? Open an issue or join the discussions on GitHub.
diff --git a/docs/index.html b/docs/index.html
index f2da683..9e5abca 100644
--- a/docs/index.html
+++ b/docs/index.html
@@ -137,6 +137,7 @@ td { color: var(--text); }
+