diff --git a/agent/credential_pool.py b/agent/credential_pool.py
index 4ccf67fe5a4a0..c12cfee279504 100644
--- a/agent/credential_pool.py
+++ b/agent/credential_pool.py
@@ -2387,17 +2387,33 @@ def _seed_from_singletons(provider: str, entries: List[PooledCredential]) -> Tup
         # env vars (COPILOT_GITHUB_TOKEN / GH_TOKEN).  They don't live in
         # the auth store or credential pool, so we resolve them here.
         try:
-            from hermes_cli.copilot_auth import resolve_copilot_token, get_copilot_api_token
+            from hermes_cli.copilot_auth import (
+                COPILOT_ENV_VARS,
+                resolve_copilot_token,
+                get_copilot_api_token,
+            )
+            # All-sources suppression gate BEFORE any work — including the
+            # `gh auth token` subprocess spawn.  resolve_copilot_token()
+            # shells out (~30ms), and the exchange retries 3x with backoff
+            # (~13s worst case); a user who suppressed every copilot source
+            # (hermes auth remove copilot gh_cli) must not pay either on
+            # every pool load (model picker open, /model, agent startup).
+            # Enumerating the full source space here matches what
+            # credential_sources._remove_copilot_gh suppresses, so an
+            # all-suppressed check is stable.
+            copilot_sources = ["gh_cli"] + [f"env:{v}" for v in COPILOT_ENV_VARS]
+            if all(_is_suppressed(provider, s) for s in copilot_sources):
+                return changed, active_sources
             token, source = resolve_copilot_token()
             if token:
                 source_name = "gh_cli" if "gh" in source.lower() else f"env:{source}"
-                # Suppression gate BEFORE the network exchange.  The
-                # exchange retries 3x with backoff (~13s worst case), so a
-                # source the user already suppressed (hermes auth remove
-                # copilot gh_cli) must not burn that dead time on every pool
-                # load (model picker open, /model, agent startup) just to
-                # have the entry discarded afterwards.  This is the same
-                # early-gate pattern every other singleton branch uses.
+                # Per-source suppression gate (a user may suppress only the
+                # gh CLI path and keep an env var, or vice versa) BEFORE the
+                # network exchange.  The exchange retries 3x with backoff
+                # (~13s worst case), so a source the user already suppressed
+                # must not burn that dead time just to have the entry
+                # discarded afterwards.  Same early-gate pattern every other
+                # singleton branch uses.
                 if _is_suppressed(provider, source_name):
                     return changed, active_sources
                 api_token, enterprise_base_url = get_copilot_api_token(token)