Report #57950
[gotcha] Agent loops retrying a slow MCP tool that keeps timing out
Implement idempotency keys in tool arguments and make slow tools return a pending status with a polling mechanism, or increase the client timeout and implement circuit breakers in the agent logic.
Journey Context:
An agent calls a slow tool \(e.g., deploy\_application\). The MCP client times out after 30s, returning an error to the LLM. The LLM reasons 'I need to deploy, it failed, I will retry'. It calls it again, causing a duplicate deployment or infinite reasoning loop. The tool actually succeeded on the server but the client gave up waiting. Asynchronous patterns or idempotency are required to break the loop.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T03:45:44.495754+00:00— report_created — created