Connect · Claude Desktop
Give Claude Desktop a memory.
The model can search your brain whenever it decides to. Nothing to configure by hand.
01 · What you get
“Supported” means three things. Here's each one for Claude Desktop.
Not possible. Its conversations live on Anthropic's servers; there is nothing on disk to read.
None. It doesn't load a local instructions file.
MCP server registered in claude_desktop_config.json.
02 · Install
One command wires Claude Desktop and every other tool on your machine.
- Run the installer. It creates your private data repo on GitHub, generates your key into the Keychain, and asks whether to start with your past AI sessions or fresh.
curl -fsSL https://raw.githubusercontent.com/rohithhn/Open-Agentic-Brain-and-Memory/main/install.sh | bashmacOS or LinuxNode 22+gitA GitHub account - It wires Claude Desktop. Claude Desktop is pull-only: the model can search your brain when it decides to, but its chats don't flow back in. That's an architectural fact, not a roadmap item.
- Check it. Restart Claude Desktop and look for the brain's tools in the tools menu, then ask it to search your memory. For a full read-only check of every tool, run
brain doctor.
03 · Questions
Connecting Claude Desktop
Do I need an account or an API key?
No. There's no account and no server of ours. The memory lives in a SQLite file on your machine, and travels as one age-encrypted file to a private GitHub repo you own.
Will it overwrite my instructions file?
No. The briefing lives inside its own BRAIN:START/END markers, and MCP wiring is additive: it never touches entries it didn't create.
Does Claude Desktop see what I did in other tools?
Yes. There's one store for every tool on the machine, and your other machines get it through brain sync and brain restore.
What platforms work?
macOS and Linux, with Node 22+, git and a GitHub account. Windows isn't supported yet.
04 · Other tools