Copilot App: create_pull_request fails with "runtime settings are not configured for this session" but the PR is created #5028

Activity

1. github-actions

   added

   triage

   on Oct 1, 2026
2. copilot-cli-bot

   added

   area:toolsBuilt-in tools: file editing, shell, search, LSP, git, and tool call behaviorBuilt-in tools: file editing, shell, search, LSP, git, and tool call behavior

   and removed

   triage

   on Oct 2, 2026
3. copilot-cli-bot

   added the

   Bug

   issue type on Oct 2, 2026
4. johnnyhaggs commented on Oct 6, 2026

   

   johnnyhaggs

   on Oct 6, 2026

   More actions

   I'm seeing the same issue on Copilot app 1.1.26 (Windows, commit 5ab79c8) with copilotd 0.9.2 on a WSL2 remote host. I compared the app logs with the remote session's events.jsonl. The results point to a cause other than the hypothesis in the issue body.

   Findings

   1. It fails for every remote-host session and none of the local sessions.

   Session location	create_pull_request calls	Result
   WSL remote host (copilotd)	7 calls across 6 sessions (2026-09-24 → 2026-10-06)	7 of 7 returned runtime settings are not configured for this session, and each PR was created
   Local (Windows) sessions	63 calls since 2026-08-19	No calls returned this error; the only failures had real causes (422 / unlinked account)

   2. The desktop app creates the PR and records success. The remote agent receives a different, failed result for the same tool call ID.

   Desktop app log (github-app.65308.log) for one call:

   19:11:37.846 remote event pump: tool lifecycle dispatch event_type=tool.user_requested tool_call_id=toolu_01L8gx… tool_name=create_pull_request delivered=true
   19:11:38.718 tools::create_pull_request: issuing GitHub pull request create request … head_ref=copilot/… base_branch=main
   19:11:41.485 tools::create_pull_request: GitHub pull request create request succeeded … pull_request_number=3
   19:11:41.501 remote event pump: translated action action_type="chat/toolCallComplete" tool_call_id=toolu_01L8gx… success=true content=[text:1]
   

   Remote host ~/.copilot/session-state/40b6bb55-…/events.jsonl for the same call:

   {"type":"tool.execution_complete","data":{"toolCallId":"toolu_01L8gx…","success":false,
    "error":{"message":"Failed to create pull request: runtime settings are not configured for this session","code":"failure"}}}

   The remote CLI's start timestamp (19:11:37.845Z) matches the app's tool.user_requested timestamp to the millisecond. Two handlers appear to run for one call:

   • The desktop app's handler, which succeeds and creates the PR.
   • A handler on the remote host, which fails right away because the remote host has no runtime settings for the session. Its result is the one the model receives.

   So the error doesn't come from a step after PR creation. The app's success result never reaches the remote session, or a remote handler's failure replaces it.

   3. This causes duplicate calls. In the same session, the agent retried create_pull_request about 1 second later (toolu_01EjHy…). The app showed success=true for that retry but made no GitHub request (no issuing GitHub pull request create request line). The remote agent again received the "runtime settings" failure.

   Suggested fix

   For remote-host sessions, route the app's create_pull_request result back to the remote session. Alternatively, don't register or run a remote-side create_pull_request handler when the desktop app owns that tool.

   I can share redacted log excerpts if needed.
5. kranthikirang commented on Oct 7, 2026

   

   kranthikirang

   on Oct 7, 2026

   More actions

   I am seeing the same problem from MS copilot studio (github copilot harness) from last 4 days. It used to work fine. Now I wonder whether this is a problem in github-copilot vs github-mcp-server