• ACTIVE
  • READY
  • FAILED
  • HUMAN

How to fine tune models with your AI agent

Connect Forefront to your agent to start training custom versions of base models. Your AI handles the job execution using your specific data.

Ask AI about this page

Short answer

How can I fine tune models using my AI agent?

Tell your agent to start a training job and give it the location of your dataset. The agent uses the Forefront connector to trigger the process on the base model. You'll get a job ID to track progress.

Training outcomes

What happens after you trigger a job.

The results of your training request.

ACTIVE

Training in progress

The agent starts the job and gives you a tracking ID. You can check back later to see if the model is ready.

READY

New model available

Once the job finishes, your agent can immediately start using the fine-tuned version for specific tasks.

FAILED

Job error report

If the dataset is bad or the job hits a limit, your agent tells you exactly what went wrong.

HUMAN

Data cleanup needed

If training fails due to bad data, you'll need to fix your files manually before trying again.

The workflow

What your AI does when you request training.

The agent translates your natural language request into a structured training command.

  1. Analyze dataset

    Your agent looks at the data you've provided to ensure it's ready for the model.

    create_fine_tune

  2. Trigger job

    The agent sends the command to the Forefront platform to start the fine-tuning process.

    create_fine_tune

  3. Monitor status

    The agent tracks the job to see if it's running, succeeded, or failed.

    create_fine_tune

  4. Confirm completion

    Your agent notifies you when the new model is live and ready for use.

    create_fine_tune

Try it

Copy these to start.

Use these prompts to kick off a training job.

Starting points

Swap out the bracketed text with your actual file paths or model names.

Forefront Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_9p3sy2z5j8vwdMtOomVYymmFfYsP4YghbaGI6v8C/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 — Forefront capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "forefront-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_9p3sy2z5j8vwdMtOomVYymmFfYsP4YghbaGI6v8C/mcp"
    }
  }
}
Copy into chat04

  • Start a fine-tuning job on the base model using my dataset at [path/to/data.jsonl].

  • Fine tune the model using the training data found in [folder_name].

  • Trigger a new training job for [model_name] with my custom dataset.

  • Run a fine-tune job using the data from [link_to_dataset].

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

Start here

Connect Forefront once, then ask.

Just link your account once. Your login stays encrypted, and you can start training models immediately through your preferred AI client.

Connect Forefront to your AI

FAQ

How the training process behaves

  • 01

    Can my AI edit my training data?

    No. The agent can only use the data you provide to trigger the job; it cannot modify your original files.

  • 02

    Does the AI choose the hyperparameters?

    The agent uses the defaults unless you explicitly tell it which settings to use in your prompt.

  • 03

    Can I cancel a job through the chat?

    No. If a job is already running, you must manage or stop it through the Forefront dashboard.

  • 04

    What happens if my dataset is too large?

    The job will fail if the data exceeds the platform's limits, and the agent will report the error to you.

  • 05

    Will the AI automatically use the new model?

    The agent can use the new model, but you have to tell it to switch to the fine-tuned version for your next prompt.

  • More questions about Forefront? The Connector page answers them. See everything the Forefront Connector can do

Connect Forefront to Claude, Cursor, ChatGPT & more