Runs: Reviewing and Comparing Agent Reliability History

A run is the permanent record of a single Critiqor observation session. Every time you finalize an observation, Critiqor produces a reliability report and saves it as a run. Over time, your run history becomes the basis for tracking whether your agent is getting more reliable — letting you compare what changed, when, and by how much.

​

What is a run?

Each time you run critiqor finalize, Critiqor creates a run from the observation session that was active. The run contains the full reliability report for that session: the trust level, primary diagnosis, recommendations, and the runtime evidence that supports them. Runs are stored locally and identified by a sequential run ID — for example, run_001, run_002, and so on. You can open, review, and reference any run at any time.

​

Listing runs

Use critiqor runs to see all completed runs at a glance. The output is a table where each row shows:

• Run ID — the unique identifier for the run (e.g., run_001)
• Timestamp — when the run was finalized
• Trust level — the High / Moderate / Low reliability signal for that session
• Diagnosis summary — a short description of the primary finding from the run

This gives you a quick view of your agent’s reliability history without opening each run individually.

​

Opening a specific run

Use critiqor dashboard run_001 to open any run’s full dashboard — substituting the run ID you want to review. The dashboard shows the complete reliability report for that session, including the evidence behind the trust level and diagnosis. To open the most recent run without specifying an ID, run critiqor dashboard with no argument.

​

Comparing runs over time

Runs are most valuable when you use them comparatively. After making changes to your agent, start a new observation session and finalize it. Then compare the new run’s trust level and primary diagnosis against earlier runs — a change in trust level or a different diagnosis is a direct signal of whether your edits had an effect. If you’re iterating on a specific problem (for example, reducing retry loops), watch whether the relevant findings appear or disappear across consecutive runs.

​

Run storage

Runs are stored locally on your machine. They are never sent anywhere without your explicit configuration — your run history stays private unless you choose otherwise.