Context infrastructure for AI-driven development
Company Brain is a persistent memory layer for AI coding agents — company standards, project architecture, feature notes, and a chained log of what every agent did, available over MCP the moment a session starts.
A four-level context hierarchy, stored centrally and served to agents on demand — so context survives past a single chat window, and travels across every repo and every agent working on the same thing.
Org-wide standards — coding conventions, architecture principles, design system. Read-only to agents; set by humans.
What a specific project does, its architecture, its expected flow. One project can span several repos.
Shared scratch space for one feature, visible to every agent — in every repo — working on it. Can span, or precede, any single project.
A chained record of what each agent did. How a fresh session — after a context-window hand-off — picks up exactly where the last one left off.
Three pieces: a backend that's the source of truth, a CLI that binds a repo to it, and a skill that tells any agent how to use what's already been pulled down.
Postgres-backed REST + MCP + OAuth server. The single source of truth for every org, project, feature, and document.
Company Brain
cb CLIcb init binds a repo to an org/project/feature and pulls its context to disk — .git/company-brain/, local, never committed.
cb init installs a skill file that Claude Code auto-discovers — it reads only the doc titles first, then the specific files it actually needs.
Bind a repo in two commands. Full setup (backend + desktop app) is in the docs.
cd your-repo
cb init