Pinecone Assistant MCP Server

Pinecone Assistant MCP Server

What is Pinecone Assistant MCP Server about?

Pinecone Assistant MCP Server provides an implementation of the Model Context Protocol (MCP) that connects to Pinecone Assistant, allowing clients to query the assistant and receive a configurable number of result items.

How to use Pinecone Assistant MCP Server?

1. Docker (recommended)

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

   • Set PINECONE_API_KEY (required).
   • Optionally set PINECONE_ASSISTANT_HOST (default: https://prod-1-data.ke.pinecone.io).
   • Adjust LOG_LEVEL if needed.

2. Building from source

   # Install Rust first
   curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
   git clone https://github.com/pinecone-io/assistant-mcp.git
   cd assistant-mcp
   cargo build --release
   ./target/release/assistant-mcp
   

   Provide the same environment variables as above.

3. Integration with Claude Desktop Add the following snippet to claude_desktop_config.json:

   {
     "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_HERE>",
           "PINECONE_ASSISTANT_HOST": "<YOUR_PINECONE_ASSISTANT_HOST_HERE>"
         }
       }
     }
   }
   

Key features of Pinecone Assistant MCP Server

• Simple Docker image for quick deployment.
• Configurable number of results returned per query.
• Supports custom Pinecone Assistant host endpoints.
• Adjustable logging verbosity via LOG_LEVEL.
• Light Rust binary for low‑overhead operation.

Use cases of Pinecone Assistant MCP Server

• AI‑augmented chat bots that need to pull relevant context from Pinecone Assistant.
• Enterprise knowledge bases where queries are resolved through Pinecone's vector search.
• Prototype development where a developer wants a ready‑to‑use MCP endpoint without building custom integration code.
• Claude Desktop extensions to enrich Claude's responses with Pinecone‑backed data.

FAQ

Q: Do I need a Pinecone account? A: Yes, you must have a Pinecone API key and an Assistant created in the Pinecone console.

Q: Can I run the server without Docker? A: Absolutely. Build the Rust source with cargo build --release and run the resulting binary, supplying the same environment variables.

Q: How do I change the default assistant host? A: Set the PINECONE_ASSISTANT_HOST environment variable to the desired endpoint when starting the container or binary.

Q: What logging levels are available? A: Use LOG_LEVEL=debug, info (default), warn, or error to control verbosity.

Q: Is the number of results configurable per request? A: The server respects the request payload's parameters; consult the MCP spec for the exact field name to adjust result count.

Pinecone Assistant MCP Server

An MCP server implementation for retrieving information from Pinecone Assistant.

Features

• Retrieves information from Pinecone Assistant
• Supports multiple results retrieval with a configurable number of results

Prerequisites

• Docker installed on your system
• Pinecone API key - obtain from the Pinecone Console
• Pinecone Assistant API host - after creating an Assistant (e.g. in Pinecone Console), you can find the host in the Assistant details page

Building with Docker

To build the Docker image:

docker build -t pinecone/assistant-mcp .

Running with Docker

Run the server with your Pinecone API key:

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

Environment Variables

• PINECONE_API_KEY (required): Your Pinecone API key
• PINECONE_ASSISTANT_HOST (optional): Pinecone Assistant API host (default: https://prod-1-data.ke.pinecone.io)
• LOG_LEVEL (optional): Logging level (default: info)

Usage with Claude Desktop

Add this to your claude_desktop_config.json:

{
  "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_HERE>",
        "PINECONE_ASSISTANT_HOST": "<YOUR_PINECONE_ASSISTANT_HOST_HERE>"
      }
    }
  }
}

Building from Source

If you prefer to build from source without Docker:

1. Make sure you have Rust installed (https://rustup.rs/)
2. Clone this repository
3. Run cargo build --release
4. The binary will be available at target/release/assistant-mcp

Testing with the inspector

export PINECONE_API_KEY=<YOUR_PINECONE_API_KEY_HERE>
export PINECONE_ASSISTANT_HOST=<YOUR_PINECONE_ASSISTANT_HOST_HERE>
# Run the inspector alone
npx @modelcontextprotocol/inspector cargo run
# Or run with Docker directly through the inspector
npx @modelcontextprotocol/inspector -- docker run -i --rm -e PINECONE_API_KEY -e PINECONE_ASSISTANT_HOST pinecone/assistant-mcp

License

This project is licensed under the terms specified in the LICENSE file.