Skip to main content
Watch: I Built an AI App Without Leaving Databricks

AI and ML platforms - official Pinecone server

Pinecone Assistant MCP server

Every Pinecone Assistant exposes a dedicated MCP endpoint that gives agents a context tool for retrieving relevant snippets from the assistant's uploaded files. It is available as a Pinecone-managed remote endpoint or as an open-source Docker container you run yourself. It is served from your own instance at https://<YOUR_PINECONE_ASSISTANT_HOST>/mcp/assistants/<YOUR_ASSISTANT_NAME> with API key auth.

Last verified 5 sources

Install

Claude Code

About Claude Code
Terminal
claude mcp add --transport http my-assistant https://<YOUR_PINECONE_ASSISTANT_HOST>/mcp/assistants/<YOUR_ASSISTANT_NAME> --header "Authorization: Bearer <YOUR_PINECONE_API_KEY>"

Cursor

About Cursor
JSON
{
     "mcpServers": {
       "pinecone-assistant": {
         "url": "https://<YOUR_PINECONE_ASSISTANT_HOST>/mcp/assistants/<YOUR_ASSISTANT_NAME>",
         "headers": {
           "Authorization": "Bearer <YOUR_PINECONE_API_KEY>"
         }
       }
     }
}

Generic

JSON
{
     "mcpServers": {
       "pinecone-assistant": {
         "command": "docker",
         "args": [
           "run",
           "-i",
           "--rm",
           "-e",
           "PINECONE_API_KEY",
           "-e",
           "PINECONE_ASSISTANT_HOST",
           "pinecone/assistant-mcp"
         ],
         "env": {
           "PINECONE_API_KEY": "<YOUR_PINECONE_API_KEY>",
           "PINECONE_ASSISTANT_HOST": "<YOUR_PINECONE_ASSISTANT_HOST>"
         }
       }
     }
}

Notes and caveats

  • Remote URL is per assistant: host and assistant name come from the Pinecone console.
  • Auth is a Pinecone API key sent as a Bearer token.
  • The docs describe a single context tool but do not name it, so no tool names are listed.
  • The old SSE endpoint (/sse suffix) is deprecated in favor of streamable HTTP.
  • The docs pull ghcr.io/pinecone-io/assistant-mcp but run the image as pinecone/assistant-mcp.
  • The Claude Desktop remote setup needs a third-party proxy (supergateway).
  • The local server is written in Rust; latest GitHub release v0.1.0 (2025-04-17).

More MCP servers for AI and ML platforms

Related reading

FAQ

What does the Pinecone Assistant MCP server do?

Every Pinecone Assistant exposes a dedicated MCP endpoint that gives agents a context tool for retrieving relevant snippets from the assistant's uploaded files. It is available as a Pinecone-managed remote endpoint or as an open-source Docker container you run yourself.

How do I add the Pinecone Assistant MCP server to Claude Code?

Run claude mcp add --transport http my-assistant https://<YOUR_PINECONE_ASSISTANT_HOST>/mcp/assistants/<YOUR_ASSISTANT_NAME> --header "Authorization: Bearer <YOUR_PINECONE_API_KEY>", copied from Pinecone's docs. This page also has snippets for Cursor and Generic.

Does the Pinecone Assistant MCP server need authentication?

Yes, API key. Transport: Streamable HTTP, stdio. Status: Unlabelled, last checked October 17, 2026.

Sources

Last verified against 5 Pinecone pages, then re-checked independently.

Newsletter

New MCP servers, the week they ship

One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.