Reflexio
Open-source learning layer that turns agent mistakes into permanent fixes
Visit website →About
Reflexio is an open-source self-improvement layer for AI agents. It records real interactions, extracts the useful lesson from a correction or a failed path, tests it, and retrieves that context before similar tasks later, so the agent stops starting from zero every time.
Learnings stay scoped to a single user by default. Patterns that recur across many users can be aggregated into shared playbooks, reviewed, and approved before they change behaviour for everyone, with no model retraining involved. Every learned behaviour is auditable and can be edited or rejected.
It installs as a lightweight SDK without rewriting the agent. The open-source package is self-hosted with local storage under an Apache 2.0 licence, and a hosted tier adds managed deployment, governance surfaces, and offline playbook optimisation.
AI-assisted draft, human-reviewed before publishing — see how we choose & review tools.
Why we picked it
Reflexio made the list for turning a correction into a permanent fix rather than a one-time apology — it extracts the actual lesson from a failed path and retrieves it before similar tasks. The scoped-then-aggregated learning model is worth naming: private by default, shared when a pattern recurs.