GBrain MCP

​

Install GBrain

1. Install the GBrain CLI:

bun install -g github:garrytan/gbrain

2. Create a local GBrain database:

gbrain init --pglite

3. Verify the install:

gbrain doctor

4. Import markdown sources you want GBrain to search:

gbrain import ~/notes

​

Use a local MCP client

1. Open your MCP client’s server configuration.
2. Add a server named gbrain.
3. Set the command to gbrain.
4. Set the arguments to serve.

{
    "gbrain": {
        "command": "gbrain",
        "args": ["serve"]
    }
}

claude mcp add gbrain -- gbrain serve

​

Use a remote MCP connector

1. Start the HTTP MCP server:

gbrain serve --http --port 3131

2. Copy the admin bootstrap token printed by the server.
3. Open the admin dashboard:

http://localhost:3131/admin

4. Paste the bootstrap token to enter the dashboard.
5. Create a client or access token for the AI client you want to connect.
6. Expose the HTTP server with your preferred deployment or tunnel.
7. In your AI client’s connector settings, add the GBrain MCP server URL:

https://YOUR-DOMAIN.example/mcp

8. If the client asks for a bearer token, create one for that client:

gbrain auth create "client-name"

9. Enter the token as a bearer token or request header:

{
    "Authorization": "Bearer YOUR_GBRAIN_ACCESS_TOKEN"
}

10. Save the connector.

​

Use GBrain in chats

1. Open a chat in the connected AI client.
2. Enable or select the gbrain connector.
3. Ask the client to search, read, or synthesize from your GBrain memory.

Use GBrain to summarize what my notes say about our MCP integration plans.

​

Connection values

Command: gbrain
Arguments: serve

Server URL: https://YOUR-DOMAIN.example/mcp
Authorization: Bearer YOUR_GBRAIN_ACCESS_TOKEN