Skip to directory

About Pinecone Developer MCP

Local Pinecone developer MCP provides Pinecone documentation search, index discovery and configuration, integrated-inference record upsert and search, cascading search, and reranking.

What you can use it for

  • Pinecone index development
Connection and setup
Hosting
Local MCP server
Transport
stdio
Endpoint
Local process; no hosted endpoint documented
Authentication
Run `@pinecone-database/mcp` locally; Pinecone API key is required for project operations, while documentation search works without a key.

Requirements

  • Node.js 20 or later
  • node and npx available in PATH
  • Pinecone project and API key for index operations
  • MCP-compatible client

Setup steps

  1. Create a Pinecone API key in the Pinecone console if database/index operations are needed; omit it only when using documentation search.
  2. For Cursor, add the official configuration to project .cursor/mcp.json or user-level ~/.cursor/mcp.json: command npx, args [-y, @pinecone-database/mcp], and optional PINECONE_API_KEY environment value.
  3. Use the client-specific configuration in the official Pinecone MCP documentation for Claude Desktop or Claude Code.
  4. Reload the client and verify with search-docs before performing index operations.
Tools 9

Pinecone official docs enumerate all nine stored tools. The server supports indexes with integrated embedding; an API key is required.

  • search-docs

    Pinecone MCP operation: search docs.

    01
  • list-indexes

    Pinecone MCP operation: list indexes.

    02
  • describe-index

    Pinecone MCP operation: describe index.

    03
  • describe-index-stats

    Pinecone MCP operation: describe index stats.

    04
  • create-index-for-model

    Pinecone MCP operation: create index for model.

    05
  • upsert-records

    Pinecone MCP operation: upsert records.

    06
  • search-records

    Pinecone MCP operation: search records.

    07
  • cascading-search

    Pinecone MCP operation: cascading search.

    08
  • rerank-documents

    Pinecone MCP operation: rerank documents.

    09
Limitations and notes
  • Only indexes using Pinecone integrated inference are supported; standalone vector indexes and Pinecone Assistants are outside this server’s scope.
Original sources
  • official tool reference

    Pinecone official docs explicitly enumerate the nine stored tools. Only indexes using integrated embedding are supported; API key required.

  • official documentation or source

    Pinecone official docs explicitly enumerate the nine stored tools. Only indexes using integrated embedding are supported; API key required.

Documentation checked 2026-10-28T12:55:35.000Z. Connection has not been runtime tested.

Similar tools