chroma-mcp

What it is and what it does

Chroma MCP Server exposes Chroma's vector database capabilities through the Model Context Protocol, a standardized interface for connecting LLM applications to external tools and data sources. It allows Claude and other MCP-compatible AI models to create and manage document collections, perform semantic and full-text search, and apply metadata filtering—all without the model needing direct database access.

The server supports multiple client modes: ephemeral (in-memory, for testing), persistent (file-based), HTTP (connecting to self-hosted Chroma instances), and cloud (connecting to Chroma Cloud). It integrates with several embedding providers including Cohere, OpenAI, and Voyage AI, and persists the chosen embedding function with each collection so that future queries use the same embeddings automatically.

Use it for:

• Add persistent memory to Claude Desktop by storing conversation context or knowledge bases in a vector database for semantic retrieval.
• Build AI-powered search over custom document collections, enabling models to find relevant context using semantic similarity rather than keyword matching.
• Integrate Chroma's vector search into LLM workflows via MCP, allowing models to query and update embeddings as part of their reasoning process.
• Test vector database functionality in development using ephemeral (in-memory) collections without external infrastructure.
• Connect Claude to a self-hosted or cloud-based Chroma instance for production workloads requiring scalability or data isolation.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Chroma MCP Server bridges Chroma's vector database with LLM applications via the Model Context Protocol, enabling AI models to create, manage, and search document collections using vector and semantic search.

Yes, if you are building LLM applications with Claude Desktop or another MCP-compatible client and need vector search and semantic memory. The low install friction and permissive license make it accessible. However, the aging maintenance status (last release over a year ago) and lack of recent updates warrant caution for production use—verify that the package's stability and feature set meet your needs before committing to it for critical workflows.

Install

chroma-mcp on PyPI

pip install chroma-mcp

uv add chroma-mcp

poetry add chroma-mcp

Installing chroma-mcp

Before you install

Low install friction with a pure-Python wheel distribution. Maintenance status is aging—last release was over a year ago (2025-08-14), though the repository remains active with recent commits (2025-09-17) and modest community engagement (585 stars).

License in practice

Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for most production and proprietary projects.

Quickstart

pip install chroma-mcp

# Configure in Claude Desktop's claude_desktop_config.json:
"chroma": {
    "command": "uvx",
    "args": ["chroma-mcp"]
}

Requires Python 3.10 or later. Intended for use as an MCP server within Claude Desktop or compatible LLM applications, not as a standalone library.

Verify before relying

• Whether embedding function persistence works correctly with collections created before Chroma v1.0.0
• Performance characteristics and scalability limits for large document collections
• Stability and breaking-change history across the aging release cycle

Package facts

Evidence: chroma_mcp-0.2.6-py3-none-any.whl

Keywords: chroma, embeddings, llm, mcp, vector-database

Tags

More Python Modules packages