Back to myrna
MCP

Notes MCP connector

The myrna MCP connector lets an MCP-compatible (Model Context Protocol) AI assistant read and write your notes. It runs locally and only needs the app address and a token — no database credentials.

1. Generate a token

In myrna, open Settings → API / MCP and click "Generate token". Copy it right away — it is shown only once. Check "Read-only" if the client should only read your notes.

2. Install the dependencies

In the project's mcp/ folder, install the dependencies:

cd mcp
npm install

3. Configure your client

Register the server in your MCP client with two variables: MYRNA_URL (the app address) and MYRNA_TOKEN (the token). Adjust the myrna-notes.mjs path to where you cloned the project. Most clients use a configuration file in the format below.

Configuration example (mcpServers format)

{
  "mcpServers": {
    "myrna-notes": {
      "command": "node",
      "args": ["/caminho/para/csNotes/mcp/myrna-notes.mjs"],
      "env": {
        "MYRNA_URL": "https://seu-myrna.app",
        "MYRNA_TOKEN": "myrna_…"
      }
    }
  }
}
If your client offers a command line to register MCP servers, point it to the same file (node …/myrna-notes.mjs) with the MYRNA_URL and MYRNA_TOKEN variables.

4. Use it

Ask your assistant things like "search my notes about X" or "create a note titled Y". A read-only token only searches and reads.

Read-only vs read + write

A read + write token creates, edits and sends notes to the trash; a read-only token only searches and reads. Give the client the least access it needs.

Security

Each token is tied to your account and only touches your notes. The token is the only secret — never share it, and revoke it anytime in Settings → API / MCP.