This project implements the Model Context Protocol (MCP) server, providing database capabilities for Chroma, an open-source embedding database. It enables LLM applications to seamlessly integrate with external data sources and tools, offering LLMs the context they need through vector search, full-text search, and metadata filtering. The target audience includes Python and JavaScript developers building LLM applications requiring persistent or ephemeral data storage and retrieval.

How It Works

The server acts as a bridge, translating LLM requests into Chroma database operations. It supports various client types (ephemeral, persistent, HTTP, cloud) and manages collections, documents, and embedding functions. This approach standardizes context provision, allowing LLMs to efficiently access and utilize external data via vector search and metadata filtering, with embedding function persistence simplifying retrieval.

Quick Start & Requirements

• Install: uvx chroma-mcp
• Prerequisites: Environment variables for API keys (e.g., CHROMA_COHERE_API_KEY) are required for specific embedding functions.
• Configuration: Can be configured via command-line arguments or environment variables (.env file support).
• Docs: Chroma MCP Docs

Highlighted Details

• Supports multiple embedding functions (Cohere, OpenAI, Jina, VoyageAI, Roboflow) with persistence.
• Offers flexible client types: ephemeral, persistent file-based, HTTP self-hosted, and Chroma Cloud.
• Provides comprehensive document and collection management operations.
• Integrates with tools like Claude Desktop for context provision.

Maintenance & Community

The project is part of the broader Chroma ecosystem. Further community and roadmap details can be found on the main Chroma project pages.

Licensing & Compatibility

The README does not explicitly state the license for chroma-mcp. The main Chroma project is typically Apache 2.0 licensed, but this specific component's license requires verification.

Limitations & Caveats

Embedding function persistence requires Chroma version 1.0.0 or later; collections created with older versions will not support this feature. API keys for external embedding functions should be managed securely, preferably via environment variables or a .env file.