Skip to main content
When you want to know exactly what an agent did, open the run.

​
Traces

Traces is the full trajectory. Every step appears in order:
  • tool calls and results
  • commands the agent ran
  • files it read or changed
  • tokens used along the way
Each step can also show the agent’s reasoning. Reading the trajectory shows whether the agent misread the task, could not find something it needed, or hit an error it could not recover from.

​
Output

Output shows the final assistant message and the recorded workspace changes from the run. Use it when you care about what the agent delivered, not only how it got there. Run Output tab with the final response and recorded workspace output.

​
Evals

Evals scores the run against your rubric: pass or fail per requirement. See Evals and Insights.

​
Export transcript

On run detail, choose Export transcript. Oqoqo downloads the recorded trajectory artifact for the run. The file type depends on the stored artifact, commonly .jsonl, .json, or .txt. From the CLI or MCP:

​
Compare runs

Select 2 or 3 runs on the Runs list and choose Compare to review Traces, Output, and Evals side by side. See Reading results. Often the fastest path is Frictions first, then jump into Traces at those points. For a plain-language walkthrough, use Ask Oqo.