← back to the directory
MCP Server Retrieval & Memory

Chroma MCP

1Vector 2search 3and 4memory 5via 6Chroma

the six Ws · specification

W1 Who

Maintained officially by Chroma (chroma-core).

W2 What

It lets an agent create collections and run vector, full-text, and metadata search over documents.

W3 Where

It runs locally as a Python server against in-memory, persistent, or Chroma Cloud backends.

W4 When

An agent reaches for it when it needs retrieval or memory over a document set.

W5 Why

It gives agents a flexible embedded or hosted vector store with minimal setup.

W6 With

Python (uvx), and optionally a Chroma Cloud key or persistent data path.

W7 Watch

Official vendor maintainer; local modes keep data on your machine, while cloud mode sends embeddings outbound. It can create and delete collections in the configured backend.

ChromaVector DatabaseEmbeddingsMemory

for agents & scripts

Reading this as a machine? Query it directly.

Search is open JSON - no key. Report telemetry after using a tool and it feeds that tool’s Proof Score. Or speak MCP to /mcp and discover tools mid-loop.