ChromaDB

You deploy a Streamable HTTP MCP server that exposes ChromaDB as a remote data store for AI assistants, enabling cross-device semantic search and vector operations while keeping your data private on your infrastructure.

How to use

Connect your MCP clients to the remote server to share a single ChromaDB instance across Claude Desktop, Claude Mobile, and Claude Code. Use the HTTP MCP endpoint for standard remote access and, if you prefer local command execution, leverage the dedicated local wrapper option. Authentication protects access when you expose the service beyond your private network.

How to install

Follow these steps to deploy the MCP server with Docker using the one-command installer or perform a manual setup.

curl -fsSL https://raw.githubusercontent.com/meloncafe/chromadb-remote-mcp/release/scripts/install.sh | bash

Configuration and security

Configure authentication, origins, and data storage to protect access and control who can query ChromaDB. The system supports multiple authentication methods and origin checks to prevent unauthorized use.

Key environment variables (in your .env) include the MCP_AUTH_TOKEN for public access protection, CHROMA_DATA_PATH for data storage, and port/origin settings to tailor exposure.

Notes on connectivity patterns

Claude Desktop and Mobile connect via a custom server URL, while Claude Code uses a CLI-based setup. All clients share the same ChromaDB instance, so embeddings and semantic search results persist across devices.

Security and access patterns

Prefer VPN access when possible and enable token-based authentication for public exposure. Regularly rotate tokens and enforce HTTPS for external access.

Troubleshooting

If the MCP server does not respond, check logs, verify the token, and confirm the MCP endpoint path is correct. For local testing, ensure the health endpoint returns a positive status.

Available tools

chroma_list_collections

List all ChromaDB collections available in the connected datastore.

chroma_create_collection

Create a new ChromaDB collection to organize documents and embeddings.

chroma_delete_collection

Delete a collection and its contents from ChromaDB.

chroma_get_collection_info

Retrieve metadata for a specific collection.

chroma_get_collection_count

Get the number of documents in a collection.

chroma_peek_collection

Preview documents within a collection.

chroma_add_documents

Add documents along with their embeddings to a collection.

chroma_query_documents

Perform semantic search against embeddings to retrieve relevant documents.

chroma_get_documents

Retrieve documents by IDs or using filters.

chroma_update_documents

Update existing documents and their metadata.

chroma_delete_documents

Delete documents from a collection.