I wanted a simple way to track my workouts. Of course, there are already plenty of fitness applications available, but I needed something more personal – a system that could choose exercises specifically for me rather than simply follow a predefined plan.
I already had a general idea of what I wanted to achieve, but I did not know how to select the right exercises or how the plan should evolve over time. I wanted the training to be adaptive: progress should be tracked, completed sessions and feedback should influence future workouts, and AI should be able to suggest changes after collecting enough data.
That idea turned into Ironman, an application built largely through vibe coding. It can effectively act as a personal trainer: it creates and presents exercises, tracks training sessions, monitors progress, and adjusts the plan after several workouts based on the collected information.
The application is also connected through MCP to an AI agent that can analyze the training data at any time. The agent describes its responsibilities as:
- managing and reviewing the training plan in Ironman,
- analyzing progress, completed sessions, and user feedback,
- identifying potential issues related to pain, technique, or excessive workload,
- proposing cautious changes to the training plan, but only after explicit user approval,
- explaining how exercises should be performed correctly,
- maintaining safe progression without making medical diagnoses or encouraging all-out training.
The result is more than a workout tracker. It is an adaptive training system where the application records what happens during workouts, while the AI helps interpret that data and gradually improve the plan.