kubeflow-trainer-mcp-server

eoinfennessy/kubeflow-trainer-mcp-server

3.1

If you are the rightful owner of kubeflow-trainer-mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

An MCP server for Kubeflow Trainer that facilitates the management of training jobs.

Kubeflow Trainer MCP Server

An MCP server for Kubeflow Trainer.

Provides tools for listing training runtimes, creating training jobs, listing training jobs, and deleting training jobs.

Usage

Follow the instructions below to configure this server for use with an MCP client.

Prerequisite

TODO (eoinfennessy): Create container image so Podman and Docker can be used instead.

1. Clone repo

git clone https://github.com/eoinfennessy/kubeflow-trainer-mcp-server.git

2. Configure MCP client

This will vary slightly with each client. Here's the mcp.json for using this with Cursor:

{
  "mcpServers": {
    "kubeflow-trainer": {
      "command": "uv",
      "args": [
        "--directory",
        "<absolute_path_to_repo_directory>",
        "run",
        "server.py"
      ]
    }
  }
}