• DATA
  • RUNS
  • METRICS
  • HUMAN

How to train computer vision models with your AI

Connect Roboflow to your AI client once. From there, you can manage datasets and training jobs through simple chat commands.

Ask AI about this page

Short answer

How can I train computer vision models with my AI?

Your AI uses Roboflow tools to fetch specific dataset versions, trigger training, and pull performance results. You can also stop or cancel active runs directly in your chat. This gives you a direct line to your training pipeline.

Training outcomes

What you get from the training loop

The AI handles the heavy lifting of the training lifecycle.

DATA

Versioned datasets

The AI pulls the exact dataset version you need. You get consistent data for every training attempt.

RUNS

Active training

Your AI kicks off training jobs on Roboflow. You can monitor progress and manage these jobs through chat.

METRICS

Performance data

The AI retrieves training results. You can immediately see how your model is performing.

HUMAN

Manual control

You decide when to stop or cancel a run. The AI executes your commands, but you keep the final say.

The workflow

What your AI does when the request arrives.

The AI acts as the bridge between your instructions and the Roboflow engine.

  1. Select data

    The AI identifies the specific dataset version you want to use for the run.

    get_version

  2. Start training

    The AI tells Roboflow to begin the training process using your selected data.

    start_training

  3. Check results

    The AI pulls the training metrics once the job is complete.

    get_training_results

  4. Manage jobs

    If a run is failing or unnecessary, the AI can halt the process.

    stop_trainingcancel_training

Try it

Copy these to start.

Use these prompts to begin interacting with your Roboflow data.

Starting points

Replace the bracketed text with your specific project names or version IDs.

Roboflow Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_3PcAB3lxgco6m9RlryrdB51L1KU7rDlnqGHOwOaP/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Roboflow capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "roboflow-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_3PcAB3lxgco6m9RlryrdB51L1KU7rDlnqGHOwOaP/mcp"
    }
  }
}
Copy into chat04

  • Get the details for dataset version [version_id] in my [project_name] project.

  • Start a new training job using version [version_id].

  • Show me the training results for the last job I started.

  • Stop the current training job for [project_name].

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Roboflow once, then ask.

Connect your account once. Your credentials stay encrypted on our side, and you can start managing models immediately.

Connect Roboflow to your AI

FAQ

How the training task behaves

  • 01

    Can the AI create new datasets?

    No. The AI can only access and manage existing versions through the available tools.

  • 02

    Will the AI delete my images?

    No. There are no tools provided to delete data or assets from your Roboflow account.

  • 03

    Can I see the training loss curves?

    Yes. The AI can retrieve training results which include the performance metrics you need.

  • 04

    Can the AI decide when a model is good enough?

    No. The AI provides the metrics, but you must review them and decide if the model meets your standards.

  • 05

    Can I stop a training job that is already running?

    Yes. You can tell your AI to stop or cancel an active training job.

  • More questions about Roboflow? See everything the Roboflow Connector can do

Connect Roboflow to Claude, Cursor, ChatGPT & more