Pinecone MCP Server

Official Pinecone MCP server. Search Pinecone docs, manage vector indexes, upsert records, run semantic search, and rerank results from your AI agent.

Developer Tools by Pinecone API Key active

GitHub Docs npm: @pinecone-database/mcp

Overview

The Pinecone Developer MCP Server is the official server from Pinecone that connects coding assistants and AI agents to your Pinecone vector database. It lets agents search Pinecone documentation to answer questions accurately, configure indexes based on application needs, generate code informed by your live index configuration and data, and test integrations through upsert and search operations against real indexes.

The server exposes nine tools spanning documentation search, index lifecycle management (list, describe, create-for-model), data operations (upsert and search records with integrated inference), and advanced retrieval features like cascading search across multiple indexes with deduplication and reranking with specialized models. It is designed for use during development, so an agent can stand up an index, load data, and iterate on retrieval quality without leaving the editor.

The server is published as the npm package @pinecone-database/mcp and is maintained by Pinecone. Note that only indexes with integrated inference are supported for data operations, and a Pinecone API key is required for anything beyond documentation search.

Tools

Tool	Description
`search-docs`	Search official Pinecone documentation to answer questions.
`list-indexes`	List all Pinecone indexes in the project.
`describe-index`	Return configuration details for a specific index.
`describe-index-stats`	Retrieve statistics for an index including record counts and namespaces.
`create-index-for-model`	Create a new index with an integrated inference (embedding) model.
`upsert-records`	Insert or update records into an index using integrated inference.
`search-records`	Query an index using text input, with optional metadata filtering and reranking.
`cascading-search`	Search across multiple indexes simultaneously with deduplication.
`rerank-documents`	Rerank a list of records or documents using a specialized reranking model.

Setup Guide

Prerequisites

Node.js v18 or later, with node and npx available on your PATH
A Pinecone API key from https://app.pinecone.io/organizations/-/keys
For data operations, indexes must use integrated inference

Installation

The server runs via npx and does not require a manual install. Add it to your MCP client configuration.

Claude Desktop config

Open Claude Desktop, go to Settings > Developer > Edit Config, then add:

{
  "mcpServers": {
    "pinecone": {
      "command": "npx",
      "args": ["-y", "@pinecone-database/mcp"],
      "env": {
        "PINECONE_API_KEY": "<your pinecone api key>"
      }
    }
  }
}

Restart Claude Desktop after saving.

Cursor

Create .cursor/mcp.json in your project root with the same JSON shown above.

Claude Code

Run claude mcp add-json and paste the server entry, or edit your MCP config directly.

Notes

Without PINECONE_API_KEY, only search-docs is available; index management and querying require the key.
Only indexes created with integrated inference work with upsert-records and search-records.

Use Cases

Generate RAG application code that is grounded in your actual Pinecone index configuration and the latest Pinecone docs.
Spin up a new integrated-inference index, upsert sample documents, and validate retrieval quality from inside your editor.
Debug retrieval issues by inspecting index stats, namespaces, and record counts without leaving the chat.
Improve search relevance by running cascading search across multiple indexes and applying rerank-documents on the results.
Answer questions about Pinecone APIs, SDKs, and best practices using search-docs against the official documentation.

Example Prompts

"List all my Pinecone indexes and show stats for the one named product-catalog."
"Create a new index called support-kb using an integrated embedding model on AWS us-east-1."
"Upsert these 20 FAQ entries into the support-kb index in the en namespace."
"Search the product-catalog index for wireless noise cancelling headphones under $200, then rerank the top 20 results."
"Use search-docs to explain how to configure metadata filtering on a Pinecone serverless index."

Pros

Official server maintained by Pinecone, published as @pinecone-database/mcp on npm.
Covers the full developer loop: docs search, index management, data upsert, query, and reranking.
Includes advanced retrieval primitives like cascading search across indexes and a dedicated rerank tool.
Simple npx-based setup with a single environment variable.

Limitations

Data tools only work with indexes that use Pinecone integrated inference; bring-your-own-embedding indexes are not fully supported.
Positioned as a developer tool, so it is not optimized for high-volume production agent workloads.
Requires a Pinecone account and API key for anything beyond reading documentation.

Alternatives

Weaviate MCP server for the Weaviate vector database.
Qdrant MCP server for Qdrant vector search.
Chroma MCP server for the Chroma embedding database.