Pinecone Developer MCP Server

Content

Pinecone Developer MCP Server is the official Model Context Protocol server from Pinecone. It lets Claude and other MCP clients connect to Pinecone projects and documentation so an assistant can inspect indexes, manage data, run searches, and look up product docs without leaving the conversation.

The server is published to npm as @pinecone-database/mcp and runs over stdio. It targets Pinecone indexes that use integrated inference, where embedding and reranking happen inside Pinecone, so the assistant works with records and text rather than raw vectors.

Source Review

• https://github.com/pinecone-io/pinecone-mcp
• https://raw.githubusercontent.com/pinecone-io/pinecone-mcp/main/README.md
• https://registry.npmjs.org/@pinecone-database/mcp
• https://docs.pinecone.io/guides/operations/mcp-server

These sources were reviewed on 2026-06-11. Prefer the live repository, README, npm package metadata, and official Pinecone documentation for current setup steps, supported tools, and any changes to index or inference behavior.

Features

• Official Pinecone MCP server distributed as the @pinecone-database/mcp npm package with a pinecone-mcp binary.
• Stdio transport launched through npx -y @pinecone-database/mcp.
• search-docs tool for searching official Pinecone documentation, which works even without an API key.
• list-indexes and describe-index tools for enumerating indexes and reading their configuration.
• describe-index-stats tool for record counts, dimensions, and namespace information.
• create-index-for-model tool for creating an index backed by an integrated inference model.
• upsert-records tool for inserting or updating records using integrated inference.
• search-records tool for querying records with filtering and reranking options.
• cascading-search tool for searching across multiple indexes with result deduplication.
• rerank-documents tool for reranking records with a specialized model.

Installation

Create a Pinecone API key from the console at app.pinecone.io, then configure an stdio MCP client to launch the server with the key in its environment:

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

Without an API key the documentation search tool still works, but index management and record tools require a valid key for the target project. Keep the key in MCP client environment configuration rather than in prompts or committed files.

Use Cases

• Ask Claude to list the indexes in a Pinecone project and describe how each one is configured.
• Check record counts, dimensions, and namespaces with index statistics before running a query.
• Create an integrated-inference index for a new retrieval workflow.
• Upsert documents or records and immediately search them with integrated inference.
• Run a cascading search across several indexes and let Pinecone deduplicate the results.
• Rerank candidate records to improve the ordering of retrieval results for a RAG pipeline.
• Look up Pinecone documentation from inside the assistant while building or debugging an integration.

Limitations

Only Pinecone indexes with integrated inference are supported. Assistants, indexes without integrated inference, standalone embeddings, and direct vector search are out of scope for this MCP server, so workflows that depend on those features need the regular Pinecone SDKs or APIs.

Safety and Privacy

Treat the configured API key as project access. The server can create indexes and upsert records, so use a least-privilege or read-only key when writes are not needed and require human review before any tool call that mutates a production index. Because the package runs through npx, it downloads and executes the published version on launch; pin a known version when reproducibility matters.

Index names, configurations, statistics, record contents, and search queries can flow to the MCP client and model provider, and records may contain customer or proprietary data. Keep the PINECONE_API_KEY out of prompts and version control, and be deliberate about what data is upserted into or returned from Pinecone indexes.

Duplicate Check

No pinecone-io/pinecone-mcp, @pinecone-database/mcp, Pinecone MCP server, or pinecone.io MCP entry was found in content/mcp or README.md. The only existing Pinecone reference is an incidental mention in a setup command, not a directory entry for this server.