As AI agents are rapidly improving, there has to be some sort of way to audit their performance so we can track their improvements over time. This is what inspired me to create critiqor: I wanted something in the background that could monitor my agent while it worked and once the agent completed the task, it would render a diagnosis of my agent's work that I could act upon. I took this basic concept and developed critiqor, an eval that audits your agent's runtime performance. You might be thinking: why can I just get an agent like codex to audit my agent's runtime? Critiqor sits patiently in the background while you're working, so your agent is undisturbed. Once you're done, you'll be redirect to the critiqor dashboard, which contains all of your historical runs. You can't lose this dashboard because it's on your local computer, so it stays with you as long as you still have critiqor installed. Let me know what use cases you've been using, and whether the improvement playbook, when applied to your agent and the same task being reran, improves the performance compared to the previous runs (I've been struggling to find a task that will trigger the improvement playbook because my runs with codex are always seemingly perfect, but there's a difference between doing a good job and a normal job, critiqor intends to turn this 'normal' job into a 'good' job.

Critiqor

Runtime Intelligence for AI Agents

Runtime Intelligence for AI Agents