Talk to your intervals.icu training data from Claude, ChatGPT, or any MCP client.

Use the AI assistant you already choose. Run icuvisor locally for control over your connector and credentials, or connect through hosted HTTPS when your client needs a public endpoint.

Install

Installs like an app.

Local mode uses one self-contained binary for macOS, Windows, and Linux. No Python, Docker, or hosted account required.

Privacy

Private by default.

In local mode, your intervals.icu API key stays on your computer and icuvisor does not relay tool calls through an icuvisor server.

AI-ready

Built for AI context.

Answers stay short and clear, with units like watts, pace, sleep score, and training load spelled out.

What to ask

Useful training questions, grounded in your own data.

Review the kilometre or mile splits from my last race.
Compare my last four threshold workouts and identify the sessions behind each comparison.
Show a deterministic race-day fitness scenario from my planned load, not a forecast.
Review this athlete's readiness before I write workouts.
Install

Download, run setup, connect your client.

Pick the build for your platform — icuvisor v1.10.0 — then run icuvisor setup and connect your AI client.

Read the full quickstart → · Release notes

Capabilities

What can it access?

icuvisor gives your assistant access to the parts of intervals.icu athletes ask about most.

  • Activities, intervals, and kilometre or mile splits
  • Fitness, power, HR, and pace curves
  • Wellness and daily readiness
  • Calendar events and structured workouts, with warnings when returned intervals.icu steps are missing or changed
  • Repeated-session comparisons with source IDs and dates in the full toolset
  • Coach rosters and custom items

Open the full tool reference →

Documentation

Find what you need.

Roadmap

Where we are.

icuvisor is in its stable local-first line. Recent work has tightened install and connector guidance, validated the Windows package path, expanded planning and reporting context, and made the core MCP server easier to reuse. The next stretch is package trust, smoother updates, and guided client setup.

Read the roadmap →