home / mcp / ragrabbit mcp server
Open Source, Self-Hosted, AI Search and LLM.txt for your website
Configuration
View docs{
"mcpServers": {
"madarco-ragrabbit": {
"command": "npx",
"args": [
"@ragrabbit/mcp",
"http://<RagRabbit install>/",
"<name of your documentation>"
]
}
}
}RagRabbit MCP Server lets you expose RagRabbit’s content indexing, LLM-based search, and chat capabilities to AI clients via a local or remote MCP interface. It enables you to retrieve pages from your documentation semantically, power AI-assisted queries, and integrate with Claude Desktop or Cursor IDE for seamless AI-assisted browsing of your content.
You configure an MCP client to connect to RagRabbit and then request document pages or run searches through the MCP channel. The MCP server exposes a command you run locally to start the integration with your RagRabbit instance. Use the provided command in your environment to start the MCP endpoint, then connect your AI clients (Claude Desktop, Cursor IDE) to that endpoint to retrieve pages and perform semantic searches.
Prerequisites you need to have before starting:
- Node.js 20.x or newer.
- PostgreSQL with pgVector enabled.
- OpenAI API Key.
- (Optional) Trigger.dev API Key for job triggering.
The MCP server is invoked using a stdio-based command that runs an MCP wrapper for RagRabbit. The explicit command shown for this setup is:
npx @ragrabbit/mcp http://<RagRabbit install>/ <name of your documentation>This starts a local MCP server that your AI clients can query. Use the base URL you provide to Claude Desktop or Cursor IDE to direct those clients to RagRabbit’s document search capabilities.
To enable Claude Desktop to use RagRabbit’s MCP server, define an MCP server in Claude Desktop with the following configuration block. This registers RagRabbit as a documentation source your AI can query.
{
"mcpServers": {
"<name_of_your_documentation_no_spaces>": {
"command": "npx",
"args": ["@ragrabbit/mcp", "http://<RagRabbit install>/", "<name of your documentation>"]
}
}
}For Cursor IDE you add a new MCP of type command with the same structure so the IDE can access RagRabbit through the MCP server.
- Chat Widget: embedable AI chat agent and instant search on your site.
- Website Crawler: RagRabbit indexes pages with PostgreSQL pgVector for fast semantic search.
- LLMs.txt Generation: generates a customizable LLM.txt with ToC reorder for your docs.
- Fumadocs integration: replace the search dialog with a custom RagRabbit component.
Set the OpenAI API key and authentication as required by your RagRabbit deployment. If you enable email-based login, configure the appropriate environment flags to control email delivery and access restrictions.
Common environment variables you may need include OPENAI_API_KEY, ADMIN_USER, ADMIN_PASSWORD, RESEND_AUTH, RESEND_ALLOWED_EMAILS, and SIMULATE_EMAILS depending on your login method.
Protect access to the MCP server with your chosen auth method. If you enable email-based login, restrict access to approved emails and consider simulating emails in development.
If you encounter issues connecting AI clients, verify the MCP command is running, the RagRabbit instance is reachable at the provided base URL, and the OpenAI API key is valid. Check logs for errors related to authentication, database connectivity, or missing environment variables.
Embeddable AI chat agent and instant search widget you can place on your site to enable interactive querying of RagRabbit content.
Crawls and indexes website content using pgVector in PostgreSQL for fast semantic search.
Generates a customizable LLM.txt with ToC reorder to tailor your document index.
Provides a command-based MCP endpoint to expose RagRabbit documentation to AI clients like Claude Desktop and Cursor IDE.
Configures Claude Desktop to access RagRabbit content through an MCP server entry.
Configures Cursor IDE to access RagRabbit content through an MCP server entry.
Replaces the search dialog with RagRabbit components for a seamless UI experience.