Troubleshooting Execution Results

Troubleshooting Execution Results

💡 Note:TL;DRIf a workflow fails, you will see it in the Execution History. Click the log to see the exact error node. The most common issues are missing API credentials or incorrect input data.

Sometimes, a workflow run will fail. Because workflows run inside a secure sandbox, we capture detailed logs so you or the Builder can figure out what went wrong.

(Remember: If a workflow crashes mid-execution, you are refunded the unused authorization amount. You only pay for what actually ran. A run that already sent its answer before a later step failed costs its full price).

Viewing the Logs

When a run is completed (or fails), you can view its results in the Execution History:

1. Go to the Executions tab in your Producer Dashboard.
2. Click on a specific run to open the Execution Results Modal.
3. Here, you will see a JSON output of the exact path the data took through the workflow.

Common Errors & Fixes

1. Missing Credentials or 401 Unauthorized

• The Problem: The workflow tried to call an external API (like OpenAI) but your API key was missing or invalid.

• The Fix: Go to your Producer Dashboard, click the Credentials (key) icon for that workflow, and ensure your keys are correctly entered and active.

• Which keys are checked: at run time only the credentials of nodes the trigger reaches are checked, so a missing key on another branch does not block this run. Turning the workflow on still lists every credential slot.

2. Invalid Input Data

• The Problem: The workflow expected specific data in the webhook payload (like "email": "test@example.com") but you provided something else.
• The Fix: Check the original documentation provided by the Builder to ensure your JSON payload matches the required schema.

3. HTTP Request Timeout

• The Problem: An individual HTTP request timed out, or the slow external API caused the overall workflow run to exceed its 5-minute wall-clock limit.
• The Fix: If the external API being called is consistently slow, you may need to ask the Builder to redesign the workflow using an asynchronous pattern.

4. Code / Logic Bug

• The Problem: The Builder's code threw an unhandled exception.
• The Fix: This is on the Builder to fix! Builders are incentivized to keep their workflows working to earn revenue. Check back later to see if they have pushed an update.

5. Cost ceiling reached or Cost Protection Triggered

• The Problem: The workflow uses AI/LLM APIs (like OpenAI) with variable costs. The Builder set a maximum API cost per run to protect you from runaway charges. This particular run exceeded that limit.
• What happened: The workflow was stopped mid-execution. You are only charged for API calls that completed before the limit was hit — not for the full ceiling amount.
• The Fix: This is a safety feature, not a bug. If this happens frequently, the Builder may need to increase the API cost ceiling for this workflow. You can also try with simpler input data that requires less AI processing.

5a. This credential only reaches <host>; the node asks for <other host>

• The Problem: A credential only reaches its own host. The node calls a different host.
• The Fix: For a Basic or Bearer credential, open the credential, add the node's host to its allowed domains and save it again. For a service with a saved host (a company domain, a server address), point the node at that host, or save a credential for the other host. No allowed domains specified means the credential was saved before allowed domains existed. Save it again with its allowed domains.

6. 403 on a webhook call

Webhook triggers expose two links. What a 403 means depends on which one the caller hit:

A path that does not match any registered link returns 404, same as an unknown link.

Execution Status Display

All executions that encounter errors display as "Failed" (red) in the Execution History, regardless of error type. Use the "Failed" filter to find all error cases quickly.

Recovering Stuck Funds

In the rare case a run never settles (a platform issue rather than a workflow issue), the funds you authorized are not lost — you can recover them yourself.

You stay in control of your own funds at every step — no support ticket needed.

Reclaim is blocked while settlement is pending. If the platform has already queued your run for the next ~5-minute settlement batch, the Reclaim button waits until that batch resolves. This prevents double-refunds.

Related

• What if a workflow breaks?
• What happens to my credentials?