Every Run leaves a full record. When something goes wrong, the record tells you which step, and usually why.
Find a Run
- Open the Loop and click the Runs tab. You can also click Runs in the sidebar to see Runs across Loops.
- Click a Run to open its detail.
Read a Run
- Highlight 1: Why it stopped
- Highlight 2: Journey, What started it, External actions, Approvals, Token usage, Activity log
- Highlight 3: A step's output
- Highlight 1: Technical details
- Highlight 2: Run summary
| Area | What it tells you |
|---|---|
| Journey | Each step in the order it ran, including jumps and revisits. |
| Output | What each step produced. |
| Technical details | Timing, usage, and other details for that step. |
| Approvals | Who was asked, who decided, and what they chose. |
| External actions | What the Run tried to do outside raia Command. |
| Why it stopped | Why the Run ended — finished, stopped by a rule, a limit, a person, or an error. |
| Report deliveries | Whether reports were sent, if you set them up. |
Troubleshoot in four steps
- Read the stop reason. On a failed Run it appears under Why it stopped. It narrows the problem to a step, a rule, a limit, or a person.
- Find the exact step. In the Journey, open the first step whose output looks wrong — not the last step.
- Compare the input and output. Did the step get what it needed? Did it follow its prompt?
- Verify facts. Check the output against the Run objective or the other system. Do not assume a result is right because the Run succeeded.
Then fix the draft, save, test with the same input, and publish.
Common problems
The Run failed before step 1 with "Stopped after the limit on external actions was reached." Check Max external actions per run on the Safety tab. If it is 0, set it to 1, click Save limits, and test again. Today a limit of 0 stops even a text-only Run before it starts. A limit of 1 does not make an action happen.
- Highlight 1: Why it stopped gives the reason.
An agent step fails because the agent is missing or not working. Open the step and check you picked the right agent, from the organization this Loop belongs to. If the agent is correct, ask an admin to check that its key and access are set up and working. Never paste keys or secrets into a prompt, a Run objective, or a screenshot.
An automatic trigger never starts a Run. Check three things: the Loop has been published, the Loop is switched on (not Paused), and the trigger itself is Switched on. Automatic triggers only start the published version.
A step adds facts that were not in the input. Tighten the prompt: "Use only… Do not add…". Limit What it can see to the previous step.
A rule never matches. Check the spelling of the named result in the step and in the rule — they must match exactly. Check the step actually returns that result in its output. Remember the first matching rule wins.
The Run stopped at a limit. Check the Run's usage in Technical details. If the work was normal, raise the limit a little. If a step was revisited again and again, fix the routing instead.
The Run waited for approval and then stopped. The time limit kept counting while it waited. Raise the time limit, or ask approvers to respond sooner.
All steps finished but the outcome failed. The outcome checks the whole Run. Open the outcome verdict to see which requirement was not met. If no verdict appears right after the Run finished, refresh the Run first. The checks use shortened evidence: if the verdict says text is missing or cut off, expand the step in the Journey and compare its full output before you change anything. If the verdict still disagrees with what you can see, ask your admin or contact support.
A scheduled Run started at the wrong time. Schedules use UTC. Recalculate for your time zone and daylight saving.
A webhook call did not start a Run. Open the trigger's Recent calls and read the refusal reason. Check the other app uses the current signing secret.
An email did not start a Run. Open the trigger's Recent messages. Check the sender is on the allowlist, and wait a few minutes if there were attachments to convert.
You are not sure an external action happened. Check the Run's external actions, then look in the other system for proof before you retry. Retries without that check can cause duplicates.
Checkpoint
You can open any Run, name the step where it went wrong, and say what you will change in the draft.
Where to go next
- Review Write clear steps to tighten prompts and rules.
- Return to the Loops overview.