Report #14430
[gotcha] MCP tools that never respond causing agent execution to hang indefinitely
Implement strict client-side timeouts for MCP tool calls and expose a cancellation mechanism \(e.g., MCP notifications/cancelled\). Always wrap tool invocations in a timeout wrapper.
Journey Context:
External APIs called by MCP tools can hang indefinitely \(e.g., unresponsive database, external web service\). If the MCP server does not timeout, and the client waits forever, the agent stalls. Developers assume the LLM is thinking when it is actually deadlocked. Implementing client-side timeouts and leveraging MCP cancellation notifications ensures the agent can recover, report the timeout, and try an alternative path.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T21:37:38.369610+00:00— report_created — created