How it works
From blank repository to shared project memory.
You bring a Git repository and the agents you already use. ContextBrain connects the knowledge between them.
- 1
Create your knowledge repository.
Make a repository on your Git provider. We recommend starting empty so ContextBrain can initialise the vault structure.
acme/engineering-brain
Private · empty repository
- 2
Connect it to a Brain.
Choose a name, connect the repository and validate access. Use the GitHub App or a supported SSH key or token.
- Brain
- Engineering
- Repository
- acme/engineering-brain
- Connection
- GitHub App
Access validated
- 3
Authorise your agent.
Install the ContextBrain plugin for Claude Code or Codex, or connect any MCP client, then choose the Brains and permissions the connection can use.
Claude Code wants to use your ContextBrain brains.
- Start in
- Acme Software / Engineering
- Limit this agent to that brain only
- Read only
- 4
Save useful context as you work.
Keep current project context, decisions, plans and session summaries in the vault.
- record_decision Payment retries
- save_plan Card updates
- save_session Retry policy agreed
- 5
Continue in the next session.
Ask your next agent to resume from the Brain. Browse the same knowledge on the web whenever you need it.
resume checkout
Current context, 2 recent sessions, 2 decisions
MCP, in plain English
The Model Context Protocol lets compatible AI tools call ContextBrain’s tools to read and update knowledge. You authorise the connection and control its access.
Several projects, one Brain
A Brain is the vault, not your application repository. Keep related project knowledge together, or use separate Brains for separate teams and access boundaries.
Make the next session a continuation.
Connect a vault. Save the context worth keeping. Pick it up with your next agent.