Skip to main content

​
What it means

Every tool declares a timeoutMs in its ToolSpec. If a call exceeds it, Aeris aborts and returns TIMEOUT. This is different from UPSTREAM_UNAVAILABLE: the upstream may have been perfectly healthy, but our call to it took too long — usually because the request scope is too broad. Retryable, but only if you also narrow the request.

​
How to fix

  • Reduce the scope: shorter date range, fewer entities, smaller page size.
  • Wait retry_after_ms (default 2000ms) before retrying the same call.
  • For write actions, reuse the original idempotency_key — a duplicate write is impossible even if the first attempt actually landed.

​
Response envelope

{
  "ok": false,
  "error_envelope": {
    "code": "TIMEOUT",
    "message": "The tool call timed out.",
    "docs_url": "https://tryaeris.ai/docs/mcp/errors/timeout",
    "retryable": true,
    "retry_after_ms": 2000,
    "details": {
      "timeout_ms": 30000
    }
  }
}
details.timeout_ms is the ceiling the tool was configured for — useful for right-sizing your retries.