chroma_mcp_server

Chroma MCP Server provides a persistent, searchable working memory layer for AI-assisted development by integrating with Chroma. It enables automated context recall, developer-managed persistence, and bidirectional linking between chats and code changes. The server exposes MCP commands to capture, retrieve, and manage context such as chat history, codebase snippets, tool sequences, and validations. You can run the server in persistent mode to keep data across restarts and configure logging and data directories via environment variables. The included Pytest plugin usage and automated learning workflows help you validate changes and promote learning with confidence. To start using it, install the package, run the server as a Python module, and point your MCP-enabled tooling to the configured server name, along with the appropriate environment settings to enable persistent storage and logging.

Prerequisites:

• Python 3.8+ (recommended latest 3.x)
• pip (Python package manager)
• Optional: virtual environment tools (venv, virtualenv)

Installation steps:

1. Create and activate a virtual environment (optional but recommended): python -m venv venv source venv/bin/activate # On Windows use: venv\Scripts\activate

2. Install the Chroma MCP Server package (basic): pip install chroma-mcp-server

3. (Optional) Install full feature set with additional embedding models and extras: pip install "chroma-mcp-server[full]"

4. Run the server using the Python module entry point shown in the documentation (see mcp_config for recommended command): python -m chroma_mcp_server

Note: If you plan to persist data, ensure you have a writable directory for CHROMA_DATA_DIR and CHROMA_LOG_DIR. The server will create necessary collections on startup if they do not exist.

Tips and common considerations:

• Use persistent storage (CHROMA_CLIENT_TYPE=persistent) to retain memory across restarts; specify CHROMA_DATA_DIR to a writable path.
• Adjust log verbosity with LOG_LEVEL and MCP_LOG_LEVEL; DEBUG provides verbose output helpful during development.
• The server supports building and managing multiple MCP collections (e.g., chat_history_v1, codebase_v1). Use the provided collection setup workflows to initialize them.
• For tool integration, point your MCP clients to the server name configured in mcp_config and ensure environment variables match your deployment environment.
• The Pytest Plugin Usage guides in the docs provide patterns for automatic validation and learning promotion workflows; leverage them to automate quality checks.
• If you encounter permission or path issues, verify that the CHROMA_DATA_DIR and CHROMA_LOG_DIR paths exist and are writable by the process.