When Something Goes Wrong
Most issues are easy to resolve, and the app typically provides guidance on next steps. This chapter covers common scenarios and solutions. Every run is preserved and can be reviewed in OBSERVE → Executions.
Find your scenario below:
“It Stopped and is Waiting for Me”
What it means: The agent reached a decision requiring human input, so it paused on purpose.
What to do: Look for a question or approval box on the run. Read it and respond — click Approve, click Reject, or type your answer. The agent continues after you reply.
This is expected. The agent pauses when human oversight is required for important decisions. See Answer an Agent When it Pauses for further details.
“It Showed an Error Message”
What it means: Something in the run failed.
What to do: Read the error card — it usually names what went wrong and suggests a fix. Common ones:
- If it mentions a missing file, attach the file and run again.
- If it mentions a login or connection, a system it needs may be down — try later or tell your administrator.
- If you’re not sure, click Copy on the error and send it to your administrator. They can see the full record of the run.
“It’s Taking a Long Time”
What it means: Some tasks take longer to complete, such as processing large documents or executing complex, multi-step workflows.
What to do:
- Let it run. You can leave the page and come back; the run keeps going and is saved under OBSERVE → Executions.
- If the run appears stalled or is no longer needed, click Stop to end it.
- Every run also has a built-in time limit (~15 minutes). If a run reaches its time limit, it stops automatically and displays a Timed-Out status. Runs cannot continue indefinitely.
“It Said It Hit a Limit”
What it means: Every run has defined limits for time, cost, and steps. When a limit is reached, the run stops automatically and preserves all results generated so far.
What to do:
- If needed, ask your administrator to increase the agent’s limits.
- Otherwise, try a smaller or more focused request.
- If you see a Budget Exceeded message, review the run first. Repeated steps or loops can quickly consume the budget.
- For more on how these limits work and who can change them, see Safety and Guardrails.
“The Result Isn’t What I Expected”
What to do:
- Review the run in OBSERVE → Executions to identify where the agent’s behavior differed from your expectations.
- Retry with a clearer, more specific request, including the desired outcome.
- If documents are involved, verify that the correct file was attached and that the agent is using the intended knowledge source. See Knowledge Bases.
- If the issue persists, share the run with the agent owner or administrator so they can adjust the agent’s configuration.
“It Won’t Run At All”
What it means: The agent must be Active before it can be run. Draft agents are still being configured and are not available for execution.
What to do: If the agent appears in BUILD → Agents but cannot be started, it is likely still a draft. Ask the owner to activate it, or use RUN → Agents, which lists only agents that are ready to run.
“It’s Not Using the Tools I Expected”
What to do:
- Clearly describe the task you want to perform. The agent selects tools based on your request.
- If the expected tool isn’t used, it may be disabled or restricted by a guardrail. Contact the agent owner to review its configuration. See Safety and Guardrails.
“I Can’t Find my Result or an Old Run”
What to do: All runs are stored in OBSERVE → Executions, with the newest listed first. Use it to find past runs, review execution details, and download results. Completed runs remain available for reference and retrieval. See Find Your Results and Download Files.
Still Stuck?
- Click the floating ? button (bottom-right) to ask an in-app help question.
- Press Ctrl+K (Windows) or ⌘K (Mac) to jump to any page.
- Or contact your administrator — they can see the full record of any run and adjust the agent.
