How-to

How to troubleshoot a failed run

When a run fails, open it from Executions to see exactly what went wrong — then fix and re-run. Here’s the flow.

1. Open Executions and click the failed run (status FAILED or TIMEOUT).
2. Read the error message and the per-step details — the failing step is marked, with its input, output, and any tools it called.
3. Fix the cause in the agent or workflow (see common causes below).
4. Click 🔁 Replay to re-run with the exact same input, and watch the fresh run.

Common causes

1. Missing tool connection — e.g. Slack/Discord/Telegram/Notion not connected in Settings → Integrations.
2. A tool call failed — check the “Tools used” details for the API’s response.
3. TIMEOUT — a run stuck over 30 minutes is auto-marked TIMEOUT; simplify the step or check the LLM/provider.
4. Bad LLM key — if using your own key (Account → BYO keys), confirm it’s valid.

Replay uses the workflow’s current definition, so fixing the workflow and replaying is the fastest way to confirm the fix.