Pinecone

What is mcp-pinecone?

mcp-pinecone is an MCP (Model Context Protocol) server designed to facilitate interaction with Pinecone, a vector database. It allows AI models, such as Claude Desktop, to perform operations like searching and uploading records to a Pinecone index, thereby enabling basic Retrieval Augmented Generation (RAG) features.

How to use mcp-pinecone?

mcp-pinecone can be installed and configured for use with Claude Desktop. You can install it via Smithery using npx -y @smithery/cli install mcp-pinecone --client claude or locally using uvx install mcp-pinecone or uv pip install mcp-pinecone. After installation, you need to configure Claude Desktop by adding mcpServers entries to its configuration file (claude_desktop_config.json), specifying the command and arguments for running mcp-pinecone. You will also need a Pinecone account, an API key, and an index name.

Key Features of mcp-pinecone

• Read and Write to Pinecone Index: Enables seamless interaction with your Pinecone index.
• Implemented Tools: Provides a set of tools for various operations:
  • semantic-search: Search for records in the Pinecone index.
  • read-document: Read a specific document from the Pinecone index.
  • list-documents: List all documents stored in the Pinecone index.
  • pinecone-stats: Get statistics about your Pinecone index (number of records, dimensions, namespaces).
  • process-document: Process documents by chunking, embedding (via Pinecone's inference API), and upserting them into the Pinecone index.
• Integration with Claude Desktop: Designed to work as an MCP server for Claude Desktop, extending its capabilities with external knowledge.

Use Cases of mcp-pinecone

• Retrieval Augmented Generation (RAG): Enhance AI model responses by retrieving relevant information from a Pinecone index.
• Knowledge Base Integration: Connect AI models to external knowledge bases stored in Pinecone for more informed responses.
• Document Management: Efficiently store, search, and manage documents within a Pinecone vector database.
• Semantic Search Applications: Build applications that leverage semantic search capabilities over your data in Pinecone.

FAQ about mcp-pinecone

Q: What is the Model Context Protocol (MCP)? A: The Model Context Protocol is a standard that allows AI models to interact with external services and data sources, extending their capabilities beyond their pre-trained knowledge.

Q: How does mcp-pinecone handle embeddings? A: mcp-pinecone generates embeddings using Pinecone's Inference API.

Q: What is uv and why is it recommended for installation? A: uv is a fast Python package installer and resolver. It is recommended for its speed and efficiency in managing Python environments.

Q: How can I debug mcp-pinecone? A: You can debug mcp-pinecone using the MCP Inspector, which can be launched via npx @modelcontextprotocol/inspector uv --directory {project_dir} run mcp-pinecone.

Pinecone Model Context Protocol Server for Claude Desktop.

Read and write to a Pinecone index.

Components

Resources

The server implements the ability to read and write to a Pinecone index.

Tools

• semantic-search: Search for records in the Pinecone index.
• read-document: Read a document from the Pinecone index.
• list-documents: List all documents in the Pinecone index.
• pinecone-stats: Get stats about the Pinecone index, including the number of records, dimensions, and namespaces.
• process-document: Process a document into chunks and upsert them into the Pinecone index. This performs the overall steps of chunking, embedding, and upserting.

Note: embeddings are generated via Pinecone's inference API and chunking is done with a token-based chunker. Written by copying a lot from langchain and debugging with Claude.

Quickstart

Installing via Smithery

To install Pinecone MCP Server for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install mcp-pinecone --client claude

Install the server

Recommend using uv to install the server locally for Claude.

uvx install mcp-pinecone

OR

uv pip install mcp-pinecone

Add your config as described below.

Claude Desktop

On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json On Windows: %APPDATA%/Claude/claude_desktop_config.json

Note: You might need to use the direct path to uv. Use which uv to find the path.

Development/Unpublished Servers Configuration

"mcpServers": {
  "mcp-pinecone": {
    "command": "uv",
    "args": [
      "--directory",
      "{project_dir}",
      "run",
      "mcp-pinecone"
    ]
  }
}

Published Servers Configuration

"mcpServers": {
  "mcp-pinecone": {
    "command": "uvx",
    "args": [
      "--index-name",
      "{your-index-name}",
      "--api-key",
      "{your-secret-api-key}",
      "mcp-pinecone"
    ]
  }
}

Sign up to Pinecone

You can sign up for a Pinecone account here.

Get an API key

Create a new index in Pinecone, replacing {your-index-name} and get an API key from the Pinecone dashboard, replacing {your-secret-api-key} in the config.

Development

Building and Publishing

To prepare the package for distribution:

1. Sync dependencies and update lockfile:

uv sync

2. Build package distributions:

uv build

This will create source and wheel distributions in the dist/ directory.

3. Publish to PyPI:

uv publish

Note: You'll need to set PyPI credentials via environment variables or command flags:

• Token: --token or UV_PUBLISH_TOKEN
• Or username/password: --username/UV_PUBLISH_USERNAME and --password/UV_PUBLISH_PASSWORD

Debugging

Since MCP servers run over stdio, debugging can be challenging. For the best debugging experience, we strongly recommend using the MCP Inspector.

You can launch the MCP Inspector via npm with this command:

npx @modelcontextprotocol/inspector uv --directory {project_dir} run mcp-pinecone

Upon launching, the Inspector will display a URL that you can access in your browser to begin debugging.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Source Code

The source code is available on GitHub.

Contributing

Send your ideas and feedback to me on Bluesky or by opening an issue.