Chroma MCP Server

Instructions

Args:
    collection_name: Name of the collection to create
    embedding_function_name: Name of the embedding function to use. Options: 'default', 'cohere', 'openai', 'jina', 'voyageai', 'ollama', 'roboflow'
    metadata: Optional metadata dict to add to the collection

Implementation Reference

• src/chroma_mcp/server.py:179-209 (handler)

  The handler function for 'chroma_create_collection' tool. It creates a new ChromaDB collection with the specified embedding function and metadata using the global Chroma client.

  @mcp.tool()
  async def chroma_create_collection(
      collection_name: str,
      embedding_function_name: str = "default",
      metadata: Dict | None = None,
  ) -> str:
      """Create a new Chroma collection with configurable HNSW parameters.
      
      Args:
          collection_name: Name of the collection to create
          embedding_function_name: Name of the embedding function to use. Options: 'default', 'cohere', 'openai', 'jina', 'voyageai', 'ollama', 'roboflow'
          metadata: Optional metadata dict to add to the collection
      """
      client = get_chroma_client()
      
      embedding_function = mcp_known_embedding_functions[embedding_function_name]
      
      configuration=CreateCollectionConfiguration(
          embedding_function=embedding_function()
      )
      
      try:
          client.create_collection(
              name=collection_name,
              configuration=configuration,
              metadata=metadata
          )
          config_msg = f" with configuration: {configuration}"
          return f"Successfully created collection {collection_name}{config_msg}"
      except Exception as e:
          raise Exception(f"Failed to create collection '{collection_name}': {str(e)}") from e

• src/chroma_mcp/server.py:171-178 (helper)

  Dictionary of known embedding functions used by the chroma_create_collection handler to select the appropriate embedding function based on the input parameter.

  mcp_known_embedding_functions: Dict[str, EmbeddingFunction] = {
      "default": DefaultEmbeddingFunction,
      "cohere": CohereEmbeddingFunction,
      "openai": OpenAIEmbeddingFunction,
      "jina": JinaEmbeddingFunction,
      "voyageai": VoyageAIEmbeddingFunction,
      "roboflow": RoboflowEmbeddingFunction,
  }