The Tredict MCP Server works with several AI assistants. The setup is similar across all of them, but the experience is not. This guide covers which client to use and why, and points you to the right setup instructions for each.

If you want to understand what the MCP Server actually gives these assistants access to, the previous article covers that in detail.

Pick Claude if you can

Structured workouts require the model to handle complex nested data across many tool calls. Claude's larger context window means it can build detailed multi-week plans with individual interval sessions without losing track of what it is building. Currently Claude is simply the best option for training plan creation, but this will hopefully change in the future as other AI assistants catch up.

The other clients handle most of the work just fine: pulling your training history, assessing fitness, answering questions about your data, and also the lighter write operations like renaming activities or editing notes. Where they fall short is generating complex nested workout plans across many tool calls without losing the thread. That is a different problem with its own nuances, worth its own post.

For most people starting out, Claude is the right choice.

The connection links, all in one place

Tredict maintains a connection guide for each supported client. The setup follows the same pattern across all of them: add the MCP Server URL, authenticate via OAuth or API key, done.

  • Connect Claude Web: best for plan creation, also reportedly works on Claude's free plan
  • Connect Claude Code: same Claude models, terminal interface, useful for long-running tasks, like complex plans, large scale activity title updates or time series comparisions
  • Connect ChatGPT: strong for reading and analysing your data
  • Connect Codex: OpenAI's coding-focused client, same read strengths
  • Connect Mistral Le Chat: EU-hosted alternative for analysis

A first prompt that tells you a lot

Once connected, a good starting prompt is asking Claude to analyse your recent training history and give an assessment of your current fitness level. That gives you a sense of what it can see and how it reasons about your data before you ask it to create anything.

The Tredict MCP Server blog post has examples of what the interaction looks like in practice.

Get your watch into the loop

The plans Claude creates only matter if they reach your wrist. Connecting a watch is a one-time setup in Tredict's device settings, not something the assistant does for you. Tredict integrates with Garmin, Coros, Suunto, Wahoo, icTrainer and others. Pick yours and authorise it. Plans that have been applied to the training calendar will sync to the watch, even if you connect it later.

Once that is done, structured workouts land on the watch automatically. From there, you just run them.

Coros Pace GPS watch on a wrist showing a workout in progress, with a wide landscape view in the background
Claude creates it, Tredict pushes it, me and the Coros Pace runs it.