MCP Pinecone Vector Database Server

What is MCP Pinecone Vector Database Server?

This server implements a Model Context Protocol (MCP) server for reading and writing vectorized information to a Pinecone vector database. It works with both RAG-processed PDF data and Confluence data, and is intended for developers using Pinecone and OpenAI embeddings in MCP‑based applications.

How to use MCP Pinecone Vector Database Server?

Install dependencies with bun install, create a .env file with your Pinecone and OpenAI API keys, then start the server with bun src/index.ts. The server listens for MCP commands via stdio. Use the example client bun examples/client.ts to test, or run the Confluence processing script bun src/scripts/process-confluence.ts <file-path> [collection] [scope].

Key features of MCP Pinecone Vector Database Server

• Search for similar documents using text queries
• Add new vectors with custom metadata
• Process and upload Confluence data in batch
• Delete vectors by ID
• Generate embeddings via OpenAI API

Use cases of MCP Pinecone Vector Database Server

• Enable semantic search over documentation exported from Confluence
• Build RAG pipelines that store and retrieve vector embeddings
• Add custom documents with metadata to a Pinecone index

FAQ from MCP Pinecone Vector Database Server

What dependencies are required?

You need the Bun runtime, a Pinecone API key, an OpenAI API key (for generating embeddings), and a configured Pinecone index.

How do I configure the Pinecone connection?

Set the environment variables PINECONE_API_KEY, PINECONE_HOST, PINECONE_INDEX_NAME, and DEFAULT_NAMESPACE in a .env file.

What transport does the server use?

The server communicates via stdio (standard input/output) for MCP commands.

Are there any temporarily disabled features?

Yes, the get-stats tool for database statistics is temporarily disabled.

Can I delete vectors from the database?

Yes, use the delete-vectors tool with a list of vector IDs and an optional namespace.