Hey Product Hunt, I built Khaos Brain because most AI memory features feel too shallow for real agent work. Saving "remember this next time" is useful, but the more valuable unit is accumulated experience: what condition appeared, what action was taken, what result happened, which route failed, and which route later became reliable. Khaos Brain is an open-source, local-first predictive memory system for AI agents. It stores experience as visible file-based cards instead of an opaque memory box. The current release is Codex-first, but the idea is broader: - before a task, the agent retrieves relevant experience - after a task, it writes observations and lessons back - maintenance workflows such as Sleep, Dream, and Architect keep the card library organized over time - organization mode can share reviewed experience models through GitHub without mixing personal preferences into the shared pool The cards are readable, searchable, reviewable, diffable, mergeable, and reversible with Git. The goal is not to replace RAG, vector databases, or ordinary notes. It is to make agent working experience visible enough that a human or team can inspect what the agent is about to reuse. Latest release: v0.4.7 Repo: https://github.com/liuyingxuvka/... Feedback discussion: https://github.com/liuyingxuvka/... The easiest way to try it is to hand the repo URL to a capable coding agent and ask it to install and enable Khaos Brain, then run the health check. The feedback I care about most: After trying it on a real AI-agent workflow, does Khaos Brain help the agent start from reusable prior experience instead of a blank context, and do visible cards with source/status/confidence feel more useful than your current AI memory, notes, or vector-store workflow?

Khaos Brain

Local predictive memory for AI agents

Local predictive memory for AI agents