remio-mcp

MCP (Model Context Protocol) server for Remio — expose your Remio knowledge base to AI agents like Claude, Cursor, and any MCP-compatible client.

中文文档

Why MCP Instead of CLI?

Remio already ships a CLI (remio search_notes, remio rag, etc.) that works great for manual use in the terminal. So why wrap it as an MCP server?

The CLI is built for humans. MCP is built for AI agents.

| | Remio CLI | remio-mcp | |---|---|---| | Usage | Manual commands in terminal | AI agent calls tools directly | | Integration | Shell scripts, task runners | Claude, Cursor, Craft Agents, any MCP client | | Structured output | --json flag required | Always structured, typed | | In Craft Agents | Requires a custom skill + Bash | Native source, works in Explore mode | | Discoverability | AI must guess command syntax | AI sees typed tool schemas |

In tools like Craft Agents, the difference is especially clear: a CLI-based skill requires the agent to compose shell commands and parse raw output every time. An MCP source is a registered, always-available integration — the agent knows exactly what tools exist, what parameters they take, and can call them in Explore (read-only) mode without extra setup.

In short: if you use Remio alongside AI agents, MCP gives you a cleaner, more reliable integration with zero friction.

Prerequisites

• Remio desktop app installed and running
• Remio CLI installed (remio in your PATH, typically ~/.local/bin/remio)
• Node.js 18+

Installation

Install via npm:

npm install -g remio-mcp

Or use directly with npx (no install needed):

npx remio-mcp

Setup

Claude Desktop / Claude Code

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "remio": {
      "command": "npx",
      "args": ["-y", "remio-mcp"],
      "env": {
        "REMIO_PATH": "/Users/YOUR_USERNAME/.local/bin/remio"
      }
    }
  }
}

Craft Agents

Add to your workspace source config (~/.craft-agent/workspaces/my-workspace/sources/remio/config.json):

{
  "name": "Remio",
  "slug": "remio",
  "type": "mcp",
  "mcp": {
    "transport": "stdio",
    "command": "npx",
    "args": ["-y", "remio-mcp"],
    "env": {
      "REMIO_PATH": "/Users/YOUR_USERNAME/.local/bin/remio"
    },
    "authType": "none"
  }
}

Available Tools

| Tool | Description | |------|-------------| | search_notes | Search notes by query, type, date, people, or collection | | read_note | Read the full content of a note by ID | | create_note | Create a new note | | update_note | Update an existing note | | delete_note | Delete a note | | add_note_to_collection | Add a note to a collection | | remove_note_from_collection | Remove a note from a collection | | create_people_note | Create a contact/people note | | rag | Ask a question using RAG over your notes |

Environment Variables

| Variable | Default | Description | |----------|---------|-------------| | REMIO_PATH | remio | Path to the remio CLI binary |

Find your remio CLI path with: which remio or ls ~/.local/bin/remio

Usage Examples

Once connected, you can ask your AI agent:

• "Search my Remio notes about design systems"
• "What do my notes say about TypeScript best practices?" (uses RAG)
• "Create a note titled 'Meeting with Alex' with these action items..."
• "Find notes from last week about the product roadmap"

Development

git clone https://github.com/GeekMai90/remio-mcp
cd remio-mcp
npm install
npm run build

Test locally:

REMIO_PATH=~/.local/bin/remio node dist/index.js

License

MIT