Khaos Brain
Open-source local memory for AI agents that stores lessons as inspectable, Git-versioned cards
Editorial take
It is best suited for developers using openai codex who want agent memory they can read, version and roll back. A free plan is available, with paid plans starting at Free (open source).
What is Khaos Brain?
Khaos Brain is an open-source (MIT) Python project that gives AI agents a local, file-based memory of past work. Instead of saving loose notes, it stores each lesson as a card: the situation, the action taken, the predicted or observed result, confidence, source, and how a future agent should use it. Cards live in a local store versioned with Git, so you can search, diff, review and roll them back. Agents read relevant cards before a task and write one observation afterward, and scheduled "Sleep" and "Dream" maintenance passes consolidate the library and test its evidence and assumptions.
The current release (v0.8.4, August 2026) is set up for OpenAI Codex out of the box, with installer-managed skills and scheduled automations; other agents can be adapted if they support pre-task retrieval, post-task write-back, local scripts and Git. An optional organization mode shares reviewed cards and skills through a private GitHub repository, and a desktop viewer lets you browse cards. (An earlier version of this page described a different product, which was incorrect.)
Best for
Developers using OpenAI Codex who want agent memory they can read, version and roll back
Score breakdown (out of 5)
What you would use it for
- →Letting a coding agent reuse lessons from earlier sessions
- →Keeping agent memory local and auditable instead of in a cloud memory service
- →Sharing reviewed engineering lessons and skills across a team through GitHub
Pros & Cons
👍 Pros
- ✓Free and MIT-licensed
- ✓Memory stays on your machine as readable files, not in a hosted service
- ✓Git history makes every change to the memory reviewable and reversible
- ✓Teams can share reviewed lessons and skills without running a memory server
👎 Cons
- ✗Set up for OpenAI Codex out of the box; other agents need adapting
- ✗Installed from source with Python; no prebuilt executable
- ✗Early-stage project (version 0.8.x) with dense, technical documentation
- ✗Assumes comfort with Git and the command line
Key Features
- ✓ Experience cards with situation, action, result, confidence and source
- ✓ Local, file-based store versioned with Git
- ✓ Pre-task retrieval and post-task write-back for agents
- ✓ Scheduled Sleep consolidation and Dream verification passes
- ✓ Optional organization mode backed by a private GitHub repository
- ✓ Desktop viewer for browsing and reviewing cards
- ✓ Installer and health check for OpenAI Codex
Available on
Integrates with
Khaos Brain Pricing
✅ Khaos Brain has a free plan - no credit card required to start.
Open source
- ✓MIT license
- ✓Runs locally with Python
- ✓Source archive from GitHub Releases (no prebuilt Windows executable)
- ✓Optional organization mode via a private GitHub repository
Developer resources
Related Tools
Make is the most visually capable automation platform available. Its canvas-based interface makes complex multi-step workflows easier to reason about than Zapier's linear format. The free tier is generous enough for light use, and paid plans offer strong value for teams automating high-volume processes.
n8n is the best automation platform for developers and technical teams who need full control over their workflows. The open-source core, self-hosting option, and code nodes make it the most extensible automation tool available. If you have hit the limits of what you can customize in Zapier, n8n is the natural upgrade.
OpenClaw is the most popular free, open-source personal AI agent. With about 390K GitHub stars and an MIT license, it handles multi-step tasks - email, calendars, browsing, files, shell commands and code - from the chat apps you already use, with no subscription. It needs technical comfort and careful permissions given its security history, but its capabilities rival paid agents.
Gumloop is the most accessible no-code automation platform for AI workflows. It has shifted from a drag-and-drop workflow builder to an AI agent platform, and connecting AI models and business tools remains practical for non-developers. For marketing and content teams who want to automate research, generation, and data enrichment without writing code, Gumloop is the right tool.
This page contains affiliate links. We may earn a commission at no extra cost to you. Learn more.