Reflexio
Turns an AI agent's corrected mistakes into rules it reuses, that you can audit.
Reflexio watches an agent's conversations, extracts corrections and failed or successful paths, and turns them into behavior rules the agent applies going forward. Each learning is revised as new sessions confirm or contradict it, and retired automatically when it goes stale — a refund-window rule from March gets replaced the moment the policy changes in June. You can open any learning, see the evidence behind it, rewrite it, approve it, or delete it, and a rejected learning stops being used immediately. It integrates via a Codex/Claude Code skill, or Python, REST, and CLI.
What holds up
- +Learnings retire automatically once newer sessions contradict them — no stale rules lingering.
- +Every learning traces back to the sessions and evidence that produced it — fully auditable.
- +Reject or delete a learning and it stops being used immediately, no redeploy needed.
Mind the limits
- −You wire the retrieve-and-publish loop into your own agent's code — no hosted plug-and-play option.
- −No pricing shown beyond a free tier — you'd need to book a demo to learn the cost.
Featured here? Take the badge
Put it on your site — it links back to this review. Free for every listed product, always.
<a href="https://stillworks.dev/products/p/reflexio/"><img src="https://stillworks.dev/badge/reflexio.svg" alt="Picked by StillWorks" width="250" height="54"></a>