cmf-mcp-server

cmf-mcp-server

A Model Context Protocol (MCP) server for CMF Server. This server exposes CMF Server functionality as tools that can be used by AI assistants.

Dependencies

Docker

• Docker install guide

Environment

Create a .env file in the project root with CMF credentials. Follow example.env

Project Structure

The project follows a modular organization for better maintainability:

/
├── main.py                # Main application entry point
├── tools/                 # MCP CMF tools organized by functionality
│   ├── __init__.py
│   ├── pipeline.py        # Pipeline-related tools
|   ├── execution.py       # Execution-related tools
|   ├── artifact.py        # Artifact-related tools
|   ├── additional.oy      # Additional / Miscellaneous tools
├── resources/             # MCP resources
│   ├── __init__.py
│   └── common.py          # Common resources (cmf://, version://)
├── prompts/               # MCP prompts
│   ├── __init__.py
│   └── guide.py           # CMF MCP Guide for LLMs
├── docker-compose.yml     # Docker Compose configuration for production
└── Dockerfile             # Docker build configuration

Running the Server

Using Docker

# Standard deployment
docker-compose up --build -d
docker-compose logs -f

The server will be available at http://localhost:8382

MCP Inspector

Can run MCP inspector using uv and the mcp CLI:

uv run mcp dev main.py

This will start an interactive interface at http://localhost:6274 where tools and resources are tested.

API

This MCP server exposes the following resources and tools. It can work across one or more CMF servers configured in .env:

Resources

• CMF://: Returns information about all configured CMF Servers
• version://: Returns the MCP server version
• list-tools://: Returns a list of tool categories
  • list-tools://{category}: Returns a list of tools within a specific category

Tools

Each tool call returns results as a list of dictionaries with the following structure:

[
  {
    "CMF": "CMF Name",
    "data": [...]  # Result data from this CMF Server
  },
  ...
]

Pipeline

• cmf_show_pipelines: Lists all Pipelines in CMF server

Execution

• cmf_show_executions: Lists all executions for a Pipeline in CMF Server
• cmf_show_execution_detail: Lists detailed executions for a Pipeline in CMF Server
• cmf_execution_lineage: Fetch execution lineage for a selected_uuid and specific_pipeline

Artifact

• cmf_show_artifact_types: Lists all artifact types in CMF Server (across all Pipelines)
• cmf_show_artifact_detail: Lists artifact details for a specific artifact_type in a specific_pipeline
• cmf_artifact_lineage: Fetch artifact lineage for a pipeline

Addtional

• cmf_show_model_card : Get model card
• cmf_show_python_env : Get Python Environment
• cmf_mlmd_pull : Fetch mlmd as a file
• cmf_mlmd_push : Push mlmd as POST body

License

• Apache 2.0 License

Usage

• To use in VSCode with properly configured copilot:
  • Add a .vscode in current project.
  • Add the following to: ./vscode/mcp.json
  • VSCode will work only with tools in MCP server. It doesn't know how to interprete resources and prompts yet

    {
      "servers": {
        "cmf-mcp-server": {
          "type": "sse",
          "url": "http://localhost:8382/sse",
          "headers": { "VERSION": "1.2" }
        }
      }
    }
    

• To use in n8n
• To use in Langflow
• To use in Claude Desktop
• To use in Cursor

Configuration/Setup

• This will read the pyproject.toml file and create requirements.txt

  pip install pip-tools
  pip-compile pyproject.toml  

• This will read the pyproject.toml file and create a uv.lock file with the locked dependencies.

  pip install uv
  uv lock