Chroma MCP Server is a self-hosted Model Context Protocol (MCP) server built on top of Chroma that provides data retrieval capabilities for LLM-powered applications. It enables AI models to create collections over generated data and user inputs, and to retrieve that data using vector search, full-text search, and metadata filtering. The server supports multiple client types: Ephemeral (in-memory) for testing, Persistent (file-based) storage, HTTP for self-hosted Chroma instances, and Cloud for Chroma Cloud integration. Collection management covers create, modify, delete, and listing with pagination; you can view collection information and statistics, configure HNSW parameters for vector search, and select an embedding function that persists with the collection. Document operations include adding documents with optional metadata and IDs, semantic search, advanced filtering by metadata or content, and retrieval by IDs or filters. Embedding functions available include default, cohere, openai, jina, voyageai, and roboflow, with per-collection persistence (not available for older versions). The server also supports environment-variable configuration, dotenv files, and CLI tooling, plus Claude Desktop integration.