Introduction
Complete customer documentation for Brain — chat, connections, workspaces, playbooks, self-hosting, and enterprise.
Brain is a self-hosted work assistant: browser chat, MCP connections, workspaces, playbooks, and schedules — running on your host with your Postgres. It is a private client with live tools, not a public Claude/ChatGPT workspace and not a RAG copy of your apps.
This manual covers:
- End users — chat, tools, playbooks, day-to-day workflows
- Self-host operators — install, configure, upgrade, keep Brain healthy
- Enterprise admins — SSO, SCIM, licenses, instance policies
Start here
| Who you are | Start with |
|---|---|
| Trying Brain locally | Local setup |
| First signed-in session | First chat |
| Deploying for your team | Self-hosting overview |
| How Brain is built | Architecture |
| Connecting Slack / GitHub / … | Connections overview |
| SSO / SCIM / licenses | Enterprise |
Self-hosted by design
Core paths do not require Vercel AI Gateway, Neon, Vercel Connect, or Vercel Sandbox. You bring Postgres, a model API key (or custom models), and a host.
What’s covered
- Getting started — local setup, short quickstart, accounts, settings map, first chat
- Chat — Ask/Agent, composer, attachments, approvals, message actions, shortcuts, projects, attached repos
- Connections — all MCP apps (OAuth, DCR, PAT, HTTP)
- Workspaces — roles, invites, shared chats
- Playbooks & schedules — saved prompts, morning brief, automatic runs
- Models — Command Code catalog and custom OpenAI-compatible endpoints (including Azure AI Foundry)
- Self-hosting — architecture, Docker, Dokku/dflow, backups, troubleshooting
- Enterprise — license, policies, instance users, SSO, SCIM
- Reference — environment variables, keyboard shortcuts, glossary, FAQ
Engineering specs (openspec/) and docs/superpowers/ plans are for contributors — not part of this site.