# brain.md

A zero-dependency CLI that gives coding agents a persistent, file-based project memory: a plain-Markdown protocol file that records durable decisions, requirements and constraints across sessions and machines.

brain.md is a toolkit, not a memory store by itself: installing it once and running an init command in a project scaffolds a BRAIN.md protocol file, a brain directory, and default configuration for agent tools already in the repository. Any coding agent that can read files, including Claude Code and Codex, then learns to use that project's brain simply by reading BRAIN.md.

The brain is meant to hold durable knowledge: the reasons behind past decisions, agreed constraints, and paths not taken, the kind of information that would otherwise live only in chat logs and vanish when a session ends. Every write goes through the brain CLI so the file's structure cannot be broken by a malformed manual edit.

It is agent-agnostic and repo-native: the brain lives in the project's own Markdown files and travels with it in git, with or without any additional runtime installed. It is distributed as an npm package and also installs skills into agent tools such as Claude Code.

Website: https://projectbrain.md
