Pinecone MCP is an MCP server that reads and writes to a Pinecone index, enabling retrieval-augmented workflows for Claude Desktop. It exposes a set of request handlers (list_resources, read_resource, list_tools, call_tool, get_prompt, list_prompts) and a collection of implemented tools (semantic-search, read-document, list-documents, pinecone-stats, process-document) that operate against a Pinecone service. Embeddings are generated via Pinecone's inference API, and documents are chunked using a token-based chunker before upserting to the index. The semantic-search flow is SemSearch -> search_records -> generate_embeddings -> index, supporting retrieval-augmented generation. Document operations flow UpsertDoc -> upsert_records, ReadDoc -> fetch_records, ListRes -> list_records. The server uses a Pinecone Client with operations including search_records, upsert_records, fetch_records, list_records, and generate_embeddings, all applied to a Pinecone Index. This MCP server is designed to integrate with Claude Desktop, with quickstart installation via Smithery or uv, and configurable for local development or published deployments. Debugging can be assisted by the MCP Inspector. The project is MIT licensed and intended to enable simple RAG features using Pinecone's Inference API.