Skill Detail

Chroma MCP Server for Embedding Database Operations

An official MCP server for the Chroma open-source embedding database. Enables AI agents to create collections, add documents, perform vector search, full-text search, and metadata filtering through the Model Context Protocol.

Integrations & ConnectorsMCP
Integrations & Connectors MCP Security Reviewed
Tool match: chroma-mcp โญ 527 GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill chroma-mcp-server-embedding-database-operations Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Last updated
Mar 26, 2026
Quick brief

The Chroma MCP Server is an official Model Context Protocol integration from the Chroma team that brings the power of the Chroma embedding database to AI assistants. Available at github.com/chroma-core/chroma-mcp, this server allows LLMs to create and manage vector collections, store documents with embeddings, and perform sophisticated retrieval operations โ€” all through MCP tool calls.

How it works

What this skill actually does

The server provides a comprehensive set of tools covering collection management (create, modify, delete, list, peek, get info, get count) and document operations (add, query, get, update, delete). Query capabilities include semantic vector search, full-text search, and advanced metadata filtering, giving AI agents flexible retrieval options depending on the use case. Collection creation supports configuring HNSW parameters for optimized vector search performance.

Multiple embedding functions are supported out of the box: default, Cohere, OpenAI, Jina, VoyageAI, and Roboflow. The server leverages Chroma’s collection configuration persistence (introduced in v1.0.0), which means once a collection is created with a specific embedding function, subsequent queries and inserts automatically use the same function without re-specification.

Deployment flexibility is a key strength. The server supports four client types: ephemeral (in-memory for testing), persistent (file-based storage), HTTP (for self-hosted Chroma instances), and cloud (automatic connection to Chroma Cloud at api.trychroma.com). Installation is as simple as running uvx chroma-mcp, and configuration is done through claude_desktop_config.json or equivalent MCP client settings. With 515+ stars and backing from the Chroma core team, this server provides a production-ready way to give AI agents long-term memory and knowledge retrieval capabilities.