Skip to directory

Chroma MCP

Chroma (official, chroma-core project)

stdio13 documented tools1

About Chroma MCP

Chroma-maintained MCP server for local, persistent, self-hosted HTTP, or Chroma Cloud vector database access. It exposes collection management, document retrieval, semantic search, and filtering tools; the chosen client mode controls where data is stored and which credentials are required.

What you can use it for

  • Create and manage vector collections
  • Add, query, retrieve, and update documents
  • Run semantic search and metadata filtering
  • Connect to Chroma Cloud or a self-hosted Chroma deployment
Connection and setup
Hosting
Local process; connects to local, self-hosted HTTP, or Chroma Cloud storage depending on client type
Transport
stdio
Endpoint
Local process; no hosted endpoint documented
Authentication
No credential for ephemeral or local persistent modes. Chroma Cloud mode requires tenant, database, and API key. Self-hosted HTTP mode may require configured custom auth credentials.

Requirements

  • Python 3.10 or later
  • uv/uvx to run the package
  • For cloud mode, Chroma Cloud tenant, database, and API key; for HTTP mode, a reachable Chroma server

Setup steps

  1. For a no-credential ephemeral local database, configure the stdio command `uvx` with args [`chroma-mcp`].
  2. Choose and configure persistent, cloud, or HTTP mode only when needed, using the documented `--client-type` and mode-specific settings.
  3. Keep cloud credentials in protected local environment configuration; do not pass real credentials in prompts or committed catalog data.
  4. Reload the MCP client, inspect the tools, then begin with a read-only collection listing or query. Review every write/delete tool before enabling destructive operations.
Tools 13

Current Chroma inventory is 13 tools; added chroma_fork_collection. Collection and document mutations affect stored vector data.

  • chroma_list_collections

    List collections with pagination support.

    01
  • chroma_create_collection

    Create a collection with optional HNSW configuration.

    02
  • chroma_peek_collection

    View sample documents from a collection.

    03
  • chroma_get_collection_info

    Get collection information and statistics.

    04
  • chroma_get_collection_count

    Get the number of documents in a collection.

    05
  • chroma_modify_collection

    Update collection name or metadata.

    06
  • chroma_delete_collection

    Delete a collection.

    07
  • chroma_add_documents

    Add documents with optional metadata and custom IDs.

    08
  • chroma_query_documents

    Query documents using semantic search and filters.

    09
  • chroma_get_documents

    Retrieve documents by IDs or filters with pagination.

    10
  • chroma_update_documents

    Update existing document content, metadata, or embeddings.

    11
  • chroma_delete_documents

    Delete documents from a collection.

    12
  • chroma_fork_collection

    Fork a Chroma collection.

    13
Limitations and notes
  • The project README labels this a beta-stage server; release behavior and available embedding backends may change.
  • Collection/document creation, modification, and deletion change stored data; start with read-only operations and confirm destructive calls.
  • External embedding providers require their own API keys; those credentials are separate from Chroma Cloud authentication.
Original sources
  • tool inventory

    Current package inventory lists 13 tools, including chroma_fork_collection.

Documentation checked 2026-10-28T00:00:00Z. Connection has not been runtime tested.

Similar tools