Chroma MCP
by Chroma
Vector search, full-text search, and metadata filtering for AI via Chroma's embedding database
database Python Intermediate Self-hostable No API key
⭐ 500 stars 📅 Updated: 1mo ago
Description
Official MCP server for Chroma, the open-source embedding database. Enables AI models to leverage vector search, full-text search, and metadata filtering for retrieval-augmented generation (RAG) and semantic memory. Manage collections, add and query documents with embeddings, update and delete entries, and perform semantic searches with advanced filtering options. Supports multiple embedding functions including Cohere, OpenAI, Jina, VoyageAI, and Roboflow, giving you flexibility in how documents are vectorized. Runs locally with no API key required, making it ideal for privacy-conscious deployments.
✅ Best for
Developers building RAG applications or semantic memory systems that need to run locally
⏭️ Skip if
You need a managed cloud vector database with built-in scaling
💡 Use cases
- Building RAG pipelines where AI retrieves relevant context from a document collection
- Creating semantic memory for AI assistants that persists across conversations
- Searching codebases or documentation by meaning rather than exact keywords
- Managing and querying large collections of embedded documents with metadata filters
👍 Pros
- ✓ No API key required — runs fully locally for maximum privacy
- ✓ Supports multiple embedding providers (Cohere, OpenAI, Jina, VoyageAI, Roboflow)
- ✓ Comprehensive collection management with 12 dedicated tools
- ✓ Combined vector search and metadata filtering for precise retrieval
👎 Cons
- ✗ Requires Python runtime and uvx for installation
- ✗ Local-only by default — no built-in cloud hosting option
- ✗ Embedding quality depends on the chosen embedding function and its configuration
🔧 Exposed tools (12 tools)
| Tool | Category | Description |
|---|---|---|
| add_documents | data | Add documents with embeddings and metadata to a collection |
| get_documents | data | Retrieve specific documents by ID or filter |
| update_documents | data | Update existing documents in a collection |
| delete_documents | data | Delete documents from a collection by ID or filter |
| list_collections | management | List all available collections in the Chroma database |
| create_collection | management | Create a new collection with specified embedding function |
| modify_collection | management | Modify collection settings and metadata |
| delete_collection | management | Delete an entire collection from the database |
| peek_collection | observation | Preview a sample of documents in a collection |
| get_collection_info | observation | Get detailed information about a collection |
| get_collection_count | observation | Get the number of documents in a collection |
| query_documents | search | Perform semantic search across documents in a collection |
⚡ Installation
Prerequisites:
- • python v3.10+
Check Claude Code documentation to configure this MCP server.
💡 Tips & tricks
Install with uvx chroma-mcp for the simplest setup. Choose your embedding function based on your use case — OpenAI embeddings are great for general text, while Roboflow is better for image-related content. Use metadata filtering to narrow down results before semantic search for best performance.
🔄 Alternatives
Quick info
- Author
- Chroma
- License
- Apache-2.0
- Runtime
- Python 3.10+
- Transport
- stdio
- Category
- database
- Difficulty
- Intermediate
- Self-hostable
- ✅
- Auth
- —
- Docker
- —
- Version
- latest
- Updated
- Feb 28, 2026
Client compatibility
- ✅ Claude Code
- ✅ Cursor
- ❓ VS Code Copilot
- ❓ Gemini CLI
- ❓ Windsurf
- ❓ Cline
- ❓ JetBrains AI
- ❓ Warp
Platforms
🍎 macOS 🐧 Linux 🪟 Windows