mcp-typesense

This MCP server exposes Typesense-backed capabilities to large language models through MCP. It provides tools to discover, query, and retrieve data from your Typesense collections, with support for schema insights and collection-level statistics. The core tools include typesense_query for full-text and filtered searches, typesense_get_document to fetch a specific document by ID, and typesense_collection_stats to retrieve metadata and counts for a collection. Prompts are available to analyze a collection’s structure (analyze_collection) and to generate effective search strategies (search_suggestions). You can use these tools by routing requests to the MCP server with the appropriate tool name, target collection, and input parameters as described in the server’s prompts and templates. The server enables zero-conf discovery by returning collection schemas and field definitions, which helps the LLM form precise queries and interpret results.

Prerequisites:

• Node.js and npm installed on your system
• Access to the MCP server repository and build setup

Install from npm (global):

npm install -g typesense-mcp-server

Install locally (project):

npm install typesense-mcp-server

Build the server (if your setup requires building TypeScript/tsc or bundling):

npm run build

Development workflow (optional):

npm install
npm run watch

Configure and run the server by pointing to your Typesense host, port, protocol, and API key. Example invocation (adjust paths and credentials as needed):

node dist/index.js \
  --host your-typesense-host \
  --port 8108 \
  --protocol http \
  --api-key your-api-key

To integrate with Claude Desktop, add an MCP server entry in your claude_desktop_config.json (see README for structure).

Notes and tips:

• The server exposes a Typesense-focused set of tools; ensure your Typesense host is reachable from where the MCP server runs.
• Use the provided templates (typesense_search, typesense_collection, etc.) to craft structured query payloads for the LLM.
• If you run into debugging issues, enable verbose logs or run the inspector tool referenced in the repository to monitor MCP communications.
• Credentials and host information should be kept secure; avoid embedding plain API keys in shared configs.
• The server supports collection discovery and returns schema details to help the model reason about data types and fields.
• Ensure your Typesense cluster permissions allow the API key to perform the requested operations (list/search/get/statistics).