Report #40018
[gotcha] Agent hangs indefinitely when MCP server tool takes too long to respond or network drops
Implement strict client-side timeouts for MCP tool calls \(e.g., 30 seconds\) and handle the timeout exception gracefully, returning a 'Tool timed out' message to the LLM so it can pivot.
Journey Context:
The MCP spec supports cancellation via JSON-RPC notifications, but if the server doesn't implement it or the client doesn't enforce it, the agent process blocks forever waiting for a response. Developers assume the server will always return quickly or fail, but external APIs flake. Client-side timeouts are mandatory to prevent zombie agents.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T21:38:38.184498+00:00— report_created — created