Chroma MCP

Official MCP server that gives LLM clients direct access to the Chroma vector database.

Free· Open source (Apache 2.0). Free to run locally or self-hosted; embedding-function API keys (OpenAI, Cohere, Jina, VoyageAI, Roboflow) billed by those providers. Chroma Cloud pricing set separately by Chroma.MCP Servers

Visit website →

Best for

Developers who already use Chroma (or want to) and need to give Claude Desktop, Cursor, or another MCP client durable vector memory or a searchable knowledge base with minimal glue code.

Skip if

Teams needing hosted, multi-tenant vector search with auth and quotas out of the box, or anyone who wants an opinionated end-to-end RAG framework rather than a database bridge.

Chroma MCP is the official Model Context Protocol server for Chroma, the open-source embedding database. It exposes Chroma's collection and document APIs as MCP tools so any MCP-aware LLM client - Claude Desktop, Cursor, Continue, Windsurf, or custom agents - can persist, search, and manage vector data without bespoke integration code. The server ships as a Python package installable via uvx or pip, and supports four client modes: ephemeral (in-memory, ideal for tests), persistent (local file-based storage), HTTP (against a self-hosted Chroma server), and cloud (Chroma Cloud at api.trychroma.com). Once wired into a client's MCP config, the LLM gets a dozen structured tools: create/list/modify/delete collections, add/get/query/update/delete documents, peek at samples, count records, and configure HNSW parameters. Query tools support semantic search plus metadata and full-text filters. Collections persist their embedding-function choice (default all-MiniLM plus optional OpenAI, Cohere, Jina, VoyageAI, or Roboflow) from v1.0.0 onward, so retrieval always uses the same vectors used at insert. Typical use: give Claude a durable long-term memory across sessions, build a shared team knowledge base an agent can search, or attach a code/document index to a coding assistant. It is the canonical way to plug Chroma into the MCP ecosystem and a good pick when you want vector search without standing up LangChain glue or writing a REST wrapper yourself.

Editor's take

The cleanest way to hand Claude a real vector store. Install is two lines of JSON, the tool surface maps 1:1 to Chroma's own API, and the ephemeral/persistent/HTTP/cloud switch means the same server scales from a laptop test to Chroma Cloud without rewriting configs. It is a bridge, not a RAG framework - bring your own chunking and re-ranking - but as an MCP citizen it is exactly what you want.

— The AI Tool Bible editorial team

Pros

✅ Official server from the Chroma team, tracks the database's features closely
✅ Four deployment modes (ephemeral, persistent, HTTP, Chroma Cloud) from one binary
✅ Twelve well-scoped MCP tools cover the full collection and document lifecycle
✅ Supports six embedding functions including OpenAI, Cohere, Jina, and VoyageAI with per-collection persistence
✅ Zero-config install via uvx - drops into Claude Desktop or Cursor in a couple of JSON lines
✅ Apache 2.0 licensed and readable Python source; easy to fork or extend
✅ Query tools expose HNSW tuning and metadata/full-text filters, not just naive semantic search

Cons

⚠️ You still have to run and back up the Chroma store yourself unless you pay for Chroma Cloud
⚠️ Command-line API key arguments are convenient but leak into process lists; env/.env path is safer but extra setup
⚠️ No built-in access control or multi-tenant isolation - anything the MCP client sees, it can delete
⚠️ Embedding-function persistence only works for collections created on Chroma v1.0.0+; older stores need migration
⚠️ Python-only server; teams on pure Node stacks add a runtime dependency
⚠️ MCP tool surface is CRUD-shaped - no higher-level RAG primitives like chunking, re-ranking, or hybrid fusion

Use cases

Long-term memory for Claude DesktopTeam knowledge base shared across agentsRAG over local documentsSemantic code search in Cursor or ContinueVector store for custom MCP agentsPersonal notes and journal recallMetadata-filtered document retrievalChroma Cloud access from LLM clients