CML

This MCP server lets you interact with Cloudera Machine Learning (CML) from a local or containerized environment. It exposes a set of MCP tools to manage projects, files, jobs, and runtime addons, enabling automation and integration with your CML workspace.

How to use

You run the MCP server locally and connect to your CML instance using an MCP client or automation script. Start the server with one of the runnable commands below, then use the available MCP tools to manage projects, files, jobs, and runtime addons. You can also configure environment credentials to keep sensitive information out of command lines.

How to install

Prerequisites you need before installing the MCP server:

• Python 3.8 or later
• pip (for Python package installation)
• an MCP client or automation tool for interacting with the server

Install the required Python packages for MCP integration.

pip install mcp[cli] requests

Optionally, you can run installation commands via the uv wrapper for isolated environments.

uv pip install mcp[cli] requests

Set up environment variables to point to your CML instance and credentials. You can use traditional environment variables or MCP configuration variables.

# Traditional environment variables
export CML_API_TOKEN="your_api_token_here"
export CML_BASE_URL="https://your-cml-instance.cloudera.com"

# MCP configuration environment variables (preferred)
export CLOUDERA_ML_API_KEY="your_api_token_here"
export CLOUDERA_ML_HOST="https://your-cml-instance.cloudera.com"

# Certificate path (optional)
export CML_CERT_FILE="/path/to/your/certificate.pem"

If you use a self-signed certificate, download the SSL certificate from your CML server and save it locally as a trusted CA bundle before starting the MCP server.

python download_certificate.py

Start the MCP server using one of the supported runtimes. The examples below show the common ways to run the script and expose the MCP endpoints.

# Using standard Python
python3 cml_mcp_server.py

# Using uv
uv run cml_mcp_server.py

# Using uvx
uvx cml_mcp_server.py

If you want to see all configuration options and usage details, run the help command after the server script is present.

python3 cml_mcp_server.py --help

You can also customize the start parameters directly, for example by providing the API key, host URL, and certificate path on the command line (when supported by the server).

python3 cml_mcp_server.py --token "your_api_token" --url "https://your-cml-instance.cloudera.com" --cert "/path/to/your/certificate.pem"

Configuration and security

The server supports two ways to configure credentials. Use MCP-specific environment variables for clean separation from application code, or rely on traditional environment variables if you prefer. Store sensitive values in secure environments and avoid embedding secrets directly in scripts.

Supported MCP environment variables (examples shown):

# MCP configuration environment variables
export CLOUDERA_ML_API_KEY="your_api_token_here"
export CLOUDERA_ML_HOST="https://your-cml-instance.cloudera.com"

# Optional certificate path for TLS validation
export CML_CERT_FILE="/path/to/your/certificate.pem"

Examples and notes

Direct usage to list projects without starting the MCP server is possible if you have a direct script or client prepared, for example to perform ad-hoc operations against your CML workspace.

Tools and capabilities you can access through the MCP server

This MCP server exposes a comprehensive set of tools to manage your CML environment. You can perform project management, file operations, job orchestration, and runtime addon discovery. See the list of core capabilities below to plan your automation.

Available tools

list_projects

List all CML projects the user has access to

create_project

Create a new CML project

get_project

Get details of a specific CML project

list_files

List files in a CML project at a specified path

read_file

Read the contents of a file from a CML project

upload_file

Upload a file to a CML project

rename_file

Rename a file in a CML project

patch_file

Update file metadata (rename, move, or change attributes)

list_jobs

List all jobs in a CML project

create_job

Create a new job in a CML project

create_job_from_file

Create a job from an existing file in a CML project

run_job

Run a job in a CML project

list_job_runs

List all runs for a job in a CML project

stop_job_run

Stop a running job in a CML project

schedule_job

Schedule a job to run periodically using a cron expression

list_runtime_addons

List all available runtime addons (e.g., Spark3, GPU)

download_ssl_cert

Download the SSL certificate from the CML server