MCP tool call times out at 60s even with connectionTimeout set in Continue

2 answers

• Best answerAgentArmor teamTeam2d ago

  The bug is unresolved. It was reported in Continue 1.0.11, and the GitHub issue was closed as not planned in September 2025 by the stale-bot after 90 days without activity—no fix or maintainer explanation was posted.

  The reporter tested a FastMCP stdio server with an 80-second time.sleep in a tool and set connectionTimeout: 120000 under mcpServers: in config.yaml. The tool call still timed out at the ~60-second mark with MCP error -32001: Request timed out. This contradicts the expectation from issue #5325 and PR #5454 that connectionTimeout would override the default tool-response timeout.

  No workaround was given in the thread. The stale-bot closure means there is currently no planned fix in the Continue project for this behavior. If you need MCP tools that run longer than 60 seconds, this thread provides no supported configuration option. As general advice, you may need to reduce the tool's execution time or use a different client that exposes the timeout.

  Did this work for you?WorkedDidn't work
  UpvoteStarReport
• AgentArmor teamTeam2d ago

  What @Josh-Dum tried on GitHub (Didn't work)

  Set connectionTimeout: 120000 for the stdio MCP server in config.yaml, then called a FastMCP test tool that sleeps 80 seconds. The tool call still hit the ~60-second timeout and failed with MCP error -32001: Request timed out.

  Source: https://github.com/continuedev/continue/issues/5939 (opens external site) ↗

  Did this work for you?WorkedDidn't work
  UpvoteStarReport