Framer AI AgentsDesign and publish professional sites with AI
Khaos Brain
AI agents that learn from past work
54 followers
AI agents that learn from past work
54 followers
Khaos Brain helps AI agents stop starting from scratch. It turns past task lessons into local, visible, Git-versioned cards, so an agent can check what worked before, what failed before, and when that experience should be trusted again.
Khaos Brain
Hey Product Hunt, I built Khaos Brain because I kept seeing the same problem with AI-agent work:
The agent may finish a task, but the useful experience from that task often disappears into chat history.
That experience is more than a note. It captures which repo rule mattered, which workflow route worked, which approach failed, what evidence made the lesson trustworthy, and whether the lesson is personal, local, public, or team-shareable.
Khaos Brain stores that as local, Git-versioned experience cards. The goal is not to create another hidden memory box. The goal is to let an agent reuse prior work in a way that a human can inspect, review, and roll back.
The current release is Codex-first because Codex can read local files, run scripts, follow repo instructions, use skills, and work with Git. Any agent with similar capabilities could adapt the same structure.
The easiest way to try it is to give the repo to a capable coding agent and ask it to install Khaos Brain, run the health check, then use it on a real task.
I would especially like feedback on the core framing: should this be described as AI memory, or is "agent work experience" closer to the real problem?
Mailwarm
Congratulations on launch! Interesting product. I will test it.