• MODEL
  • STATUS
  • DATA
  • HUMAN

How to tune foundation models with your AI

Connect your agent to IBM watsonx to run tuning jobs on your own data. Once connected, your agent handles the heavy lifting of job management and status tracking.

Ask AI about this page

Short answer

How can I tune foundation models using my AI?

You can trigger new tuning jobs and check their progress directly through your AI agent. This allows you to refine model performance on specific datasets without leaving your chat interface. You'll get real-time updates on whether a job is still running or finished.

The tuning outcomes

What you get from tuning

Running these jobs changes how your models behave on your specific data.

MODEL

Refined performance

The agent starts a tuning job that adapts a foundation model to your unique dataset. You get a model that understands your specific terminology and patterns.

STATUS

Live job tracking

Your agent monitors the progress of active tuning tasks. You'll know exactly when a model is ready for use.

DATA

Dataset alignment

The agent uses your provided datasets to guide the tuning process. This ensures the resulting model aligns with your specific domain requirements.

HUMAN

Manual oversight

You decide which datasets are ready for tuning and when to deploy the resulting model. The AI manages the execution, but you maintain control over the training data.

The workflow

What your AI does when you request tuning.

Your agent acts as the bridge between your chat interface and the IBM watsonx engine.

  1. Initiate tuning

    The agent sends your dataset and configuration to IBM watsonx to begin the refinement process.

    start_model_tuning

  2. Monitor progress

    The agent periodically checks the status of the running job to see if it is still processing.

    get_tuning_status

  3. Report completion

    Once the job finishes, the agent notifies you that the new model version is available.

    get_tuning_status

  4. Verify results

    The agent provides the current state of the tuning job so you can decide on the next steps.

    get_tuning_status

Try it

Copy these to start.

Use these prompts to kick off the process.

Starting points

Swap out the bracketed text with your actual dataset names or model IDs.

IBM watsonx Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_gXytFT8fNy6ELfNWb5ZhGBzw4TcaHPZqjIMO9o5b/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 — IBM watsonx capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "ibm-watsonx-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_gXytFT8fNy6ELfNWb5ZhGBzw4TcaHPZqjIMO9o5b/mcp"
    }
  }
}
Copy into chat04

  • Start a model tuning job using the dataset 'customer_support_v2'.

  • Check the status of my current tuning job.

  • Run a tuning job for the foundation model in my watsonx project.

  • Is my tuning job finished yet?

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

Start here

Connect IBM watsonx once, then ask.

Just link your account once. Your credentials stay encrypted, and you can immediately start managing tuning jobs through your agent.

Connect IBM watsonx to your AI

FAQ

How the tuning task behaves

  • 01

    Can the AI upload the dataset for me?

    No. You must provide the dataset location or reference so the agent can point the tuning job to the right data.

  • 02

    Can the AI delete my existing models?

    No. The agent can only start new tuning jobs and check their status; it cannot delete or modify existing models.

  • 03

    How do I know if a tuning job failed?

    The agent uses the status tool to check the job. If the job fails, the agent will report the status returned by watsonx.

  • 04

    Can I change the tuning parameters mid-job?

    No. Once a job is started, the parameters are set. You would need to start a new job with different settings.

  • 05

    Does the AI handle the actual training compute?

    No. The agent triggers the job in IBM watsonx, which handles all the heavy computational work.

  • More questions about IBM watsonx? See everything the IBM watsonx Connector can do

Connect IBM watsonx to Claude, Cursor, ChatGPT & more