pinecone-assistant

The Pinecone Assistant MCP server allows you to integrate Pinecone's powerful vector database and assistant capabilities with various AI agents and applications. It provides a standardized Model Context Protocol (MCP) interface for retrieving context snippets, building assistants, and enabling advanced AI workflows.

What it does:

• Provides an MCP endpoint for Pinecone Assistant.
• Facilitates retrieval of relevant context snippets for AI models.
• Enables integration with popular AI frameworks and tools like LangChain, Claude Desktop, and Cursor.

How to use: There are two primary ways to use the Pinecone Assistant MCP server:

1. Remote MCP Server (SSE/Streamable HTTP): Connect directly to your hosted Pinecone Assistant instance via HTTP. The endpoint URL follows the format: https://<YOUR_PINECONE_ASSISTANT_HOST>/mcp/assistants/<YOUR_ASSISTANT_NAME>. It supports both Streamable HTTP and SSE transports. Example LangChain integration:

   from langchain_mcp_adapters.client import MultiServerMCPClient
   # ... (other imports and setup)
   async with MultiServerMCPClient(
       {
           "assistant_ai_news": {
               "url": "https://prod-1-data.ke.pinecone.io/mcp/assistants/ai-news",
               "transport": "streamable_http",
               "headers": {"Authorization": f"Bearer {pinecone_api_key}"}
           }
       }
   ) as client:
       # ... (use client.get_tools() with your agent)
   

2. Local MCP Server (Stdio via Docker): Run the Pinecone Assistant MCP server locally using Docker. This provides a stdio interface. To start the server:

   docker run -i --rm \
     -e PINECONE_API_KEY=<PINECONE_API_KEY> \
     -e PINECONE_ASSISTANT_HOST=<PINECONE_ASSISTANT_HOST> \
     pinecone/assistant-mcp
   

   You can configure this in your mcp.json for tools like Claude Desktop or Cursor:

   {
     "mcpServers": {
       "pinecone-assistant": {
         "command": "docker",
         "args": [
           "run", "-i", "--rm", "-e", "PINECONE_API_KEY", "-e", "PINECONE_ASSISTANT_HOST", "pinecone/assistant-mcp"
         ],
         "env": {
           "PINECONE_API_KEY": "<YOUR_PINECONE_API_KEY>",
           "PINECONE_ASSISTANT_HOST": "<YOUR_PINECONE_ASSISTANT_HOST>"
         }
       }
     }
   }
   

Implementation Details:

• The remote server leverages Pinecone's cloud infrastructure.
• The local server is provided as a Docker image (pinecone/assistant-mcp) for easy deployment and integration.