NoteCove NoteCove

Connect Any AI Agent to NoteCove

In this tutorial you will connect an AI agent to NoteCove. This works with Claude Code, opencode, Cursor, or any tool that can execute shell commands or speak MCP.

The core idea is simple: if your agent can run binaries, just tell it about the notecove CLI and it will figure out the rest.

Before you start

You need:

  • NoteCove desktop app installed and running
  • The NoteCove CLI installed (Tools → CLI → Install CLI… in the desktop app)
  • An AI agent that can execute shell commands or connect to MCP servers

Option A: Agent can run shell commands

This is the simplest path. If your agent can execute binaries (Claude Code, opencode, Cursor terminal, etc.), all you need to do is:

  1. Authenticate in the directory where your agent runs:

    notecove init \
  --profile <your-profile-id> \
  --tasks-project <your-project-slug> \
  --notes

    Approve the request in the NoteCove desktop app.

  2. Tell the agent it has access to notecove. For example, add to your project instructions or system prompt:

    You have access to the notecove CLI for managing notes and tasks. Run notecove --help to see available commands. Use --json for machine-readable output.

That is it. The agent will discover the CLI, run --help to learn the commands, and use it naturally.

Finding your profile ID

notecove profiles

Copy the long ID string from the output.

Option B: Agent speaks MCP (stdio)

If your agent supports the Model Context Protocol, connect it to NoteCove’s MCP server in stdio mode:

  1. Set up authentication as in Option A

  2. Get the MCP config:

    notecove mcp --print-config

  3. Add to your agent’s MCP config. The format varies by agent — here are common examples:

    Claude Code (.mcp.json in project root):

    {
  "mcpServers": {
    "notecove": {
      "command": "/path/to/notecove",
      "args": ["mcp"]
    }
  }
}

    Claude Code (via CLI):

    claude mcp add notecove /path/to/notecove -- mcp

    Other MCP clients: Use the JSON from notecove mcp --print-config and adapt it to your client’s configuration format.

Option C: Agent connects over HTTP

For agents running on remote machines or in containers, use port mode:

# Start the MCP server on a port
notecove mcp --port 3456

# Get the config with bearer token
notecove mcp --print-config --port 3456

Then configure your agent to connect to http://127.0.0.1:3456/mcp with the bearer token from the config.

For remote machines, forward the port over SSH:

ssh -L 3456:localhost:3456 user@remotehost

See the MCP Server reference for details on port mode and authentication.

Verify the connection

Ask your agent to run:

notecove folder list

If it returns your folder list, you are connected.

What the agent can do

Once connected, your agent can:

  • Read and search notes: notecove note search "topic", notecove note show <id>
  • Create and edit notes: notecove note create "Title" --folder "Work"
  • Manage tasks: notecove task create, notecove task change, notecove task list
  • Organize: notecove folder create, notecove folder list
  • Comment: notecove task comments add, notecove note comments add

All commands support --json for structured output that agents can parse reliably.

Next steps