AgentFixboard
UnverifiedLast checked OpenClaw#transport#error-messages

A timed-out MCP session DELETE fails the next turn with 'MCP runtime cleanup could not confirm closure'

After a remote Streamable HTTP MCP server times out during turn preparation and its session DELETE does not finish within 5 s, the next turn in that session fails before inference with MCP runtime cleanup could not confirm closure.

From GitHub

openclaw/openclaw

IssuePull request

Difficulty
Easy
Time
~5 min
Steps
2
Fixed in
—
Last checked

Unverified: we could not confirm from the sources that a fix shipped. Source: PR #182189 merged. Last checked 2026-10-24.

Symptom

After a remote Streamable HTTP MCP server times out during turn preparation and its session DELETE does not finish within 5 s, the next turn in that session fails before inference with MCP runtime cleanup could not confirm closure.

Evidence what you'll see

  • The first turn still answers. The log can show failed to start server and MCP tool listing timed out after 10000ms, then later unused server cleanup failed: Error: MCP runtime cleanup could not confirm closure.
  • The next message fails in under a second: lane task error: … error="MCP runtime cleanup could not confirm closure". The user sees Your request couldn't be completed: MCP runtime cleanup could not confirm closure.
  • When the same server answers DELETE with 200, the following turn does not fail that way.

Why it happens

A timed-out remote session DELETE is treated as unconfirmed local cleanup. That sticky failure is rethrown when the next turn tries to replace the runtime, so the turn dies before a model request. Pull request 182189 treats remote termination as best effort after the local client has closed.

Affected versions

OpenClaw 2026.10.5 (1b58cb9). The report cites the same path on 2026.10.6 and on main c055a4968318. macOS 27.0.1 arm64, Node 26.7.0, npm global, launchd Gateway. Two mcp.servers entries use transport: "streamable-http".

Fix steps

0 of 2 done

Your check marks stay in this browser only.

How to check it worked

  • After a remote MCP tools/list timeout whose DELETE does not finish, the next turn in the same session starts instead of failing with MCP runtime cleanup could not confirm closure.

If it still fails

  • No published release tag contains pull request 182189. Tag v2026.10.7 was compared against merge 791621be9e05 and does not contain it (diverged).
  • The linked issue does not claim a later turn that still fails after local cleanup itself failed.

Sources and credit

Summarized in our own words from these public reports. Credit goes to their authors.

Provided as is, without warranty. Try changes in a non-production environment and back up first. Security-related steps: verify against the official docs. Terms · Request removal or attribution change

Hit an error we don't cover? Submit it