Skip to main content
This page documents every Critiqor release — new features, user-visible improvements, bug fixes, and breaking changes. Entries are listed in reverse chronological order so the latest changes are always at the top.
Check your installed version with pip show critiqor. Update to the latest release with pip install --upgrade critiqor.

​
v0.2.0

Current release

​
New

  • critiqor check command — validate your agent deployment against a policy file before shipping. Pass a policy file with --policy <file> to codify your acceptance criteria and block deployments that fall below your defined thresholds.
  • Visibility control — pass --visibility to critiqor monitor openclaw to set the visibility level of the monitoring session.
  • Difficulty tier selection — pass --difficulty-tier to critiqor monitor openclaw to choose the evaluation intensity: easy, standard, hard, or stress.
  • Custom OpenClaw command — use --openclaw-command to supply a custom command when launching OpenClaw, giving you full control over how the target agent process is started.
  • Custom working directory — use --cwd to specify the working directory in which OpenClaw is launched.
  • Session timeout — use --timeout to set a maximum duration for an observation session, after which the session is automatically finalized.

​
Improved

  • Observation session workflow — the end-to-end flow for starting, running, and finalizing sessions is more reliable and easier to follow.
  • critiqor runs listing — each entry in the run list now includes a diagnosis summary, so you can triage results at a glance without opening the full dashboard.
  • Dashboard reliability — general stability and reliability improvements to the dashboard experience.

​
v0.1.0

Initial release — 2026-05-29

​
New

  • critiqor monitor openclaw — launch OpenClaw and begin a runtime observation session. Critiqor watches your agent in real time and captures behavioral evidence throughout the run.
  • critiqor finalize — stop the active observation session and open the reliability dashboard for that run.
  • critiqor dashboard [run_id] — open the dashboard for the latest completed run, or pass a specific run_id to review any historical run.
  • critiqor runs — list all completed evaluation sessions stored locally.
  • Trust levels — every run is assigned one of three trust levels: High, Moderate, or Low, giving you an at-a-glance signal of agent reliability.
  • Dashboard sections — the run dashboard includes six sections: Executive Summary, Trust Assessment, Primary Diagnosis, Runtime Evidence, Recommendations, and Historical Runs.
  • Local run storage — all run artifacts are stored on your own machine. No data is sent to external servers.
  • Python support — compatible with Python 3.10, 3.11, 3.12, and 3.13.