MCP/Chroma Vector DB

Chroma Vector DB

Connects AI coding agents to Chroma vector database for semantic search, collection management, and embedding retrieval across local and cloud stores.

rag-data•Databases & Storage

GitHub600

License:Apache-2.0

VERIFIED

@Chroma

About this Server

The official Chroma Model Context Protocol (MCP) server provides seamless vector database and retrieval capabilities to AI models and coding agents. It enables agents to manage collections, insert documents with metadata and embeddings, perform semantic vector searches, full-text queries, and metadata filtering. Supports in-memory ephemeral testing, persistent local directories, self-hosted HTTP endpoints, and managed Chroma Cloud instances.

Real-time Interaction

Direct low-latency RPC protocol communication with structured JSON schemas.

Scoped Sandboxing

Restricted capability bounds to prevent accidental modifications or leakage.

Security & Best Practices

Local Profile Isolation: When using persistent storage, restrict data directories to dedicated local project storage.

API Key Protection: Store Chroma Cloud API keys in client environment variables or secure credential storage.

Data Sanitization: Ensure sensitive credentials or PII are not embedded into unencrypted vector collections.

When to Use (and When Not To)

Best For

•Building semantic search, knowledge bases, and RAG pipelines for AI assistants
•Storing and retrieving code snippets, documentation chunks, and conversational memory
•Running fast local or in-memory vector databases without external cloud dependencies

Not For / Limitations

•Relational SQL transactions and complex multi-table joins (use PostgreSQL, Neon, or Supabase MCP instead)
•Direct live website scraping or web crawling (use Firecrawl MCP instead)
•Real-time team chat messaging (use Slack MCP instead)

Install & Setup

One-line terminal command

terminal command

$claude mcp add chroma -- uvx chroma-mcp

Exposed Tools & Endpoints

12 tools

Functions your AI agent can invoke via this Model Context Protocol server

chroma_list_collections• callable tool

Lists all available Chroma collections with pagination support.

chroma_create_collection• callable tool

Creates a new collection with optional embedding function and HNSW distance configurations.

chroma_peek_collection• callable tool

Views a sample of documents and embeddings stored in a specified collection.

chroma_get_collection_info• callable tool

Retrieves detailed metadata, embedding function info, and configuration of a collection.

chroma_get_collection_count• callable tool

Returns the total number of documents currently stored in a collection.

chroma_modify_collection• callable tool

Updates the name or configuration metadata of an existing collection.

chroma_delete_collection• callable tool

Permanently deletes a Chroma collection and all its associated vector embeddings.

chroma_add_documents• callable tool

Adds documents, metadata dictionaries, and optional custom IDs to a collection.

chroma_query_documents• callable tool

Performs semantic vector search across documents with optional metadata and content filters.

chroma_get_documents• callable tool

Retrieves documents by exact IDs or metadata filters with pagination support.

chroma_update_documents• callable tool

Updates document contents, metadata, or embedding vectors for existing document IDs.

chroma_delete_documents• callable tool

Deletes specific documents from a collection by ID or filter criteria.