Assistant MCP

What pinecone-io/assistant-mcp MCP server does

The pinecone-io/assistant-mcp MCP server gives an MCP-compatible agent access to information retrieved from Pinecone Assistant. Its primary capability is knowledge retrieval: an agent can send a request through MCP and receive relevant information from the configured Assistant.

The server also supports retrieving multiple results. The number of returned results can be configured, which makes it suitable for workflows that need more than one piece of context from the Assistant rather than a single match.

This project is a focused bridge between MCP clients and Pinecone Assistant. It does not describe tools for creating Assistants, managing documents, or administering Pinecone resources. Use it when the required operation is retrieving Assistant-backed knowledge for an agent.

How it works

At runtime, the server authenticates with a Pinecone API key and sends retrieval requests to a Pinecone Assistant API host. The host is supplied through configuration; if it is not set, the server uses https://prod-1-data.ke.pinecone.io as the default.

The repository provides a Docker-based runtime and a Rust build path. With Docker, the image is built from the repository and then started as an interactive container. The MCP process runs inside that container and receives its Pinecone configuration through environment variables.

The README also documents use with Claude Desktop. Claude Desktop starts the Docker container using the configured environment values, allowing the client to discover and call the server through its MCP connection.

Setup and configuration

Docker is required for the documented container workflow. You need a Pinecone API key from the Pinecone Console and the API host shown in the details for the Assistant you created. PINECONE_API_KEY is required. PINECONE_ASSISTANT_HOST is optional because a default host is provided. LOG_LEVEL is also optional and defaults to info.

The documented image is built with:

docker build -t pinecone/assistant-mcp .

The container is then run with the required credentials and, when needed, a custom Assistant host. Keep the API key in the runtime environment rather than embedding it in source code or image configuration.

For a non-Docker build, the repository requires Rust. Running cargo build --release produces the target/release/assistant-mcp binary. The README also shows the MCP Inspector as a testing option, but it is a separate debugging utility rather than part of the server itself.

Tools and capabilities

The documented capabilities are:

• Retrieve information from Pinecone Assistant.
• Request multiple retrieval results.
• Configure the number of results returned.
• Run through Docker or a compiled Rust binary.
• Connect to Claude Desktop through its MCP server configuration.

The supplied material does not name individual MCP tool identifiers or specify the exact configuration mechanism for changing the result count. Those details should be confirmed from the server implementation before building an integration that depends on particular tool names or parameters.

Limitations and notes

The pinecone-io/assistant-mcp MCP server requires access to Pinecone credentials and an Assistant endpoint. The API key is mandatory, so deployments must provide a valid Pinecone credential. A custom Assistant host may be necessary when the Assistant is not reachable through the documented default.

The README does not state a software license name; it only refers to the repository's LICENSE file. It also does not describe Pinecone service pricing, supported Pinecone regions, response schemas, or error behavior. The documented client example covers Claude Desktop, while compatibility with other MCP clients is not established by the supplied material.