Set up Exomem

Exomem is in private beta. Setup takes about an hour, most of it waiting for platform exports.

1. Install on your computer

You need macOS or Linux, Node.js 22, pnpm, Git and either Docker or PostgreSQL 16. On Apple Silicon the embedding model runs locally; elsewhere it runs on the CPU.

git clone https://github.com/WilliamNurmi/exomem.git
cd exomem
pnpm project:setup
pnpm project:doctor

With Claude Code or Codex you can instead install the Exomem plugin and run its setup skill, which does these steps for you and asks before anything that needs your accounts.

2. Import your memories

pnpm sources:status lists supported sources and where their exports go; pnpm sources:guide <source> explains one export step by step. Then run the matching pnpm ingest:… command. Every imported source is reachable from your assistants; pnpm exposure:set <source> --local-only hides one from exomem.io.

3. Link your computer

pnpm link:pair
pnpm link:install-service --load

Approve the code on exomem.io/link. The computer must stay awake and online for your assistants to reach it.

4. Connect your assistants

• ChatGPT: install Exomem from the app directory, or add a connector with https://mcp.exomem.io/mcp.
• Claude (web, desktop, mobile): Settings → Connectors → Add custom connector → https://mcp.exomem.io/mcp.
• Claude Code: claude mcp add --transport http exomem https://mcp.exomem.io/mcp
• Codex: codex mcp add exomem --url https://mcp.exomem.io/mcp, then codex mcp login exomem.

Each one opens an exomem.io sign-in and asks you to approve the connection.