Ragie Model Context Protocol Server

What is Ragie Model Context Protocol Server about?

Provides a Model Context Protocol (MCP) server that lets AI models query a Ragie knowledge base. The server runs on Node.js, reads MCP messages from stdio, and returns matching content chunks.

How to use Ragie Model Context Protocol Server?

1. Set the API key: RAGIE_API_KEY=your_api_key.
2. Run the server with npx:

   RAGIE_API_KEY=your_api_key npx @ragieai/mcp-server
   

3. Optional flags:
   • --description, -d <text> – custom tool description.
   • --partition, -p <id> – target a specific Ragie partition.
4. Integrate with tools like Cursor or Claude Desktop by adding an MCP config (mcp.json or claude_desktop_config.json) that points to the npx @ragieai/mcp-server command and supplies the API key.

Key features of Ragie Model Context Protocol Server

• Single retrieve tool with parameters query, topK (default 8), rerank (default true), and recencyBias (default false).
• Returns an array of content chunks directly from the Ragie knowledge base.
• Supports custom tool description and partition targeting via CLI options.
• Written in TypeScript, leveraging @modelcontextprotocol/sdk, ragie, and zod for validation.
• Easy installation and execution with npx – no global install required.

Use cases of Ragie Model Context Protocol Server

• Augmented AI assistants that need up‑to‑date company knowledge while generating responses.
• Context‑aware chatbots retrieving relevant documents, policies, or FAQs on demand.
• RAG (Retrieval‑Augmented Generation) pipelines where a language model queries a curated knowledge base before generation.
• Internal support tools that surface the most relevant internal articles or code snippets.

FAQ from the Ragie Model Context Protocol Server

Q: What Node.js version is required? A: Node.js ≥ 18.

Q: Do I need to build the project before running? A: No. The package can be executed directly via npx.

Q: How do I specify a different Ragie partition? A: Use the --partition (or -p) CLI flag.

Q: Can I change the tool description shown to the model? A: Yes, with the --description (or -d) flag.

Q: What environment variable must be set? A: RAGIE_API_KEY – your Ragie authentication key.

Q: How do I integrate with Cursor? A: Add an mcp.json file in the project (.cursor/mcp.json) or globally (~/.cursor/mcp.json) containing the server configuration.

Q: Is the server compatible with Claude Desktop? A: Yes – place a claude_desktop_config.json in the appropriate application data folder with the same configuration.

Ragie Model Context Protocol Server

A Model Context Protocol (MCP) server that provides access to Ragie's knowledge base retrieval capabilities.

Description

This server implements the Model Context Protocol to enable AI models to retrieve information from a Ragie knowledge base. It provides a single tool called "retrieve" that allows querying the knowledge base for relevant information.

Prerequisites

• Node.js >= 18
• A Ragie API key

Installation

The server requires the following environment variable:

• RAGIE_API_KEY (required): Your Ragie API authentication key

The server will start and listen on stdio for MCP protocol messages.

Install and run the server with npx:

RAGIE_API_KEY=your_api_key npx @ragieai/mcp-server

Command Line Options

The server supports the following command line options:

• --description, -d <text>: Override the default tool description with custom text
• --partition, -p <id>: Specify the Ragie partition ID to query

Examples:

# With custom description
RAGIE_API_KEY=your_api_key npx @ragieai/mcp-server --description "Search the company knowledge base for information"

# With partition specified
RAGIE_API_KEY=your_api_key npx @ragieai/mcp-server --partition your_partition_id

# Using both options
RAGIE_API_KEY=your_api_key npx @ragieai/mcp-server --description "Search the company knowledge base" --partition your_partition_id

Cursor Configuration

To use this MCP server with Cursor:

Option 1: Create an MCP configuration file

1. Save a file called mcp.json

• For tools specific to a project, create a .cursor/mcp.json file in your project directory. This allows you to define MCP servers that are only available within that specific project.
• For tools that you want to use across all projects, create a ~/.cursor/mcp.json file in your home directory. This makes MCP servers available in all your Cursor workspaces.

Example mcp.json:

{
  "mcpServers": {
    "ragie": {
      "command": "npx",
      "args": [
        "-y",
        "@ragieai/mcp-server",
        "--partition",
        "optional_partition_id"
      ],
      "env": {
        "RAGIE_API_KEY": "your_api_key"
      }
    }
  }
}

Option 2: Use a shell script

1. Save a file called ragie-mcp.sh on your system:

#!/usr/bin/env bash

export RAGIE_API_KEY="your_api_key"

npx -y @ragieai/mcp-server --partition optional_partition_id

2. Give the file execute permissions: chmod +x ragie-mcp.sh

3. Add the MCP server script by going to Settings -> Cursor Settings -> MCP Servers in the Cursor UI.

Replace your_api_key with your actual Ragie API key and optionally set the partition ID if needed.

Claude Desktop Configuration

To use this MCP server with Claude desktop:

1. Create the MCP config file claude_desktop_config.json:

• For MacOS: Use ~/Library/Application Support/Claude/claude_desktop_config.json
• For Windows: Use %APPDATA%/Claude/claude_desktop_config.json

Example claude_desktop_config.json:

{
  "mcpServers": {
    "ragie": {
      "command": "npx",
      "args": [
        "-y",
        "@ragieai/mcp-server",
        "--partition",
        "optional_partition_id"
      ],
      "env": {
        "RAGIE_API_KEY": "your_api_key"
      }
    }
  }
}

Replace your_api_key with your actual Ragie API key and optionally set the partition ID if needed.

2. Restart Claude desktop for the changes to take effect.

The Ragie retrieval tool will now be available in your Claude desktop conversations.

Features

Retrieve Tool

The server provides a retrieve tool that can be used to search the knowledge base. It accepts the following parameters:

• query (string): The search query to find relevant information
• topK (number, optional, default: 8): The maximum number of results to return
• rerank (boolean, optional, default: true): Whether to try and find only the most relevant information
• recencyBias (boolean, optional, default: false): Whether to favor results towards more recent information

The tool returns:

• An array of content chunks containing matching text from the knowledge base

Development

This project is written in TypeScript and uses the following main dependencies:

• @modelcontextprotocol/sdk: For implementing the MCP server
• ragie: For interacting with the Ragie API
• zod: For runtime type validation

Development setup

Running the server in dev mode:

RAGIE_API_KEY=your_api_key npm run dev -- --partition optional_partition_id

Building the project:

npm run build

License

MIT License - See LICENSE.txt for details.