ragie

The Ragie MCP server exposes a single tool named retrieve, which lets AI models query Ragie's knowledge base via the MCP protocol. Once running, clients can connect to this MCP server and invoke the retrieve tool with a query and optional parameters such as topK, rerank, and recencyBias to control result relevance and freshness. The tool returns an array of knowledge chunks that match the query, enabling downstream agents (e.g., ChatGPT, Claude, or other MCP-enabled models) to surface relevant Ragie content in responses. Typical usage involves sending a request with a query string and tuning topK to balance coverage and latency.

To use this server in your workflow, configure an MCP client (for example via Cursor or Claude Desktop) to point at the ragie MCP server, and then call the retrieve tool with your search terms. You can customize results with options: topK (default 8), rerank (default true), and recencyBias (default false). Provide your Ragie API key as an environment variable (RAGIE_API_KEY) when starting the server to authenticate with Ragie's API.

Prerequisites:

• Node.js v18 or newer
• A Ragie API key

Installation and startup:

1. Ensure Node.js is installed and your Ragie API key is ready.

2. Start the MCP server using npx and supply your Ragie API key:

   RAGIE_API_KEY=your_api_key npx -y @ragieai/mcp-server

3. (Optional) If you need to restrict or seed the server with a partition, add the --partition flag:

   RAGIE_API_KEY=your_api_key npx -y @ragieai/mcp-server --partition your_partition_id

4. The server will listen on stdio for MCP protocol messages and can be consumed by MCP clients such as Cursor or Claude Desktop.

Environment variables and configuration tips:

• RAGIE_API_KEY is required to authenticate with Ragie's API.
• You can pass an optional --partition to target a specific Ragie partition.
• When integrating with clients like Cursor or Claude Desktop, you can create an mcp.json config or shell script as shown in the README to automate startup.
• If you encounter connection or authentication issues, verify that the RAGIE_API_KEY is correct and that network access to Ragie's API endpoints is allowed from your environment.
• The server implements a single tool named retrieve; ensure your MCP client invokes that tool with the appropriate query and parameters.