Pinecone MCP Server – Vector Search and Indexes
Pinecone's official Developer MCP Server gives AI coding assistants structured access to Pinecone documentation and supported index operations. Use it to create integrated-inference indexes, inspect index state, upsert records, run semantic searches, and rerank results from a local MCP-compatible client.
Overview
Pinecone's official Developer MCP Server connects compatible AI coding
assistants to Pinecone projects and documentation. It gives agents current
product guidance and structured access to supported index operations, reducing
reliance on static model knowledge or manually copied console information.
What the MCP server enables
The server exposes tools for Pinecone documentation, index discovery, data
ingestion, retrieval, and reranking. With a valid Pinecone API key, an AI agent
can:
- Search the official Pinecone documentation.
- List indexes and inspect their configuration.
- Retrieve index statistics, including record counts and namespaces.
- Create a new index that uses an integrated embedding model.
- Insert or update records in an integrated-inference index.
- Search records from natural-language text with metadata filtering and reranking.
- Search across multiple indexes with deduplication and reranking.
- Rerank documents or text with a Pinecone reranking model.
Documentation search remains available without a Pinecone API key, but index
management and query tools require authenticated project access.
When to use it
Use Pinecone MCP when an AI agent needs current Pinecone documentation or must
perform approved development operations against a Pinecone project. Typical
workflows include creating an index for a RAG application, inspecting existing
namespaces, loading sample documents, evaluating semantic search quality,
comparing results across indexes, and validating Pinecone SDK code against
current official guidance.
Connection and authentication
Pinecone distributes the official server as the
@pinecone-database/mcp npm package. The documented configuration launches it
locally over stdio with npx -y @pinecone-database/mcp.
Set PINECONE_API_KEY to a Pinecone API key when project and index tools are
required. Node.js 18 or later is required. The local process connects to
Pinecone's APIs directly and no hosted Streamable HTTP or SSE endpoint is
documented for this Developer MCP Server.
Key considerations
The server supports only indexes with integrated inference. Indexes that depend
on externally generated vectors, standalone embedding workflows, and Pinecone
Assistants are not supported by this server. Pinecone provides a separate
Assistant MCP server for retrieving context from Assistant knowledge bases.
Keep API keys outside source control, use the least privilege available, review
index creation and upsert actions before execution, and confirm that the target
index uses an integrated embedding model before invoking data tools.
Supported Transports
stdio
Command: npx
Args:
-y@pinecone-database/mcp
Frequently Asked Questions
- When should an AI agent use the Pinecone MCP server?
- Use it when a workflow needs current Pinecone documentation or approved operations for integrated-inference indexes, such as creating an index, inspecting index state, upserting records, running semantic search, searching across indexes, or reranking results.
- What does the Pinecone MCP server add to an AI agent's capabilities?
- It gives the agent structured access to official Pinecone documentation and live Pinecone project tools, allowing it to inspect and operate supported indexes instead of relying only on static model knowledge or manually copied console data.
- What can an AI agent access or manage through Pinecone MCP?
- The agent can search Pinecone documentation, list and describe indexes, inspect index statistics, create integrated-inference indexes, upsert records, search records, run cascading search across multiple indexes, and rerank documents or text.
- How is authentication configured for the Pinecone MCP server?
- Set PINECONE_API_KEY to a valid Pinecone API key for project, index, upsert, and search tools. Documentation search can work without the key. Store the API key in a protected environment or secret manager and never commit it to source control.
- Which transport should be used for the Pinecone MCP server?
- Use stdio with the official @pinecone-database/mcp npm package launched through npx. Pinecone's official Developer MCP documentation does not provide a hosted Streamable HTTP or SSE endpoint for this server.