Use this the moment a workflow fails silently, or replies with something unexpected, instead of re-triggering it repeatedly and hoping to spot the issue.
Steps
- Open the workflow, click the Executions tab at the top of the editor.
- Find the failed run (shown in red, with an "Error" label and how long it took).
- Click into it, the canvas re-renders showing exactly which node has a red error icon.
- Click that node directly, look for a detailed error panel, this often includes the exact request body sent and the exact response received, not just a one-line summary.
- Fix the specific node, save, republish, then either wait for a real trigger or use Execute step on that node to test in isolation before re-running the whole chain.
Gotcha
The short error message shown on the canvas ("Bad request," "Problem in node X") is often too vague to fix from alone. Always open the full error detail panel, specifically the actual request payload that was sent, before assuming you know what's wrong. More than once, the real cause (a field silently containing the wrong data type, or a value that wasn't actually being replaced by an expression) only becomes obvious once you see the literal request body.
Quick reference
Debugging order that actually works:
- Read the full error detail, not just the summary
- Isolate the failing node with "Execute step"
- Fix one thing at a time
- Republish before retesting