After much work, Tredict is pleased to introduce the Tredict MCP server.
AI-powered assistants and agents vastly expand the scope of training analysis and planning in endurance sports. As a manufacturer-independent platform, Tredict connects sports watch manufacturers such as Garmin, Suunto, Wahoo or Coros on one side, which can automatically record completed training sessions, execute structured workouts from the calendar and synchronise with Tredict, with AI-powered training analysis and training plan creation directly for Tredict through LLMs such as Claude, ChatGPT, Mistral or Gemini on the other. This makes Tredict the first independent platform to officially close this loop!
The best part, however, is that you decide whether you want to connect the Tredict MCP server to an AI provider or not. You use Tredict with the AI and Large Language Model (LLM) of your choice, not the other way around. This way Tredict stays true to its principles, and the sovereignty and security of your data is and remains sacrosanct.
A further innovation are the so-called MCP apps of the Tredict MCP server. These are interactive UI elements that can be embedded directly into Claude.ai or ChatGPT chats, extending the classic, purely text-based approach of MCP with something visual and tactile.
If you would like to learn more about the technical architecture of the Model Context Protocol, you can read the FAQ entry here:
What is the Model Context Protocol - MCP?
Mistral Le Chat sets better titles for training sessions in Tredict.
Significance for athletes and endurance sports enthusiasts
LLMs can help you with tasks and analyses that were previously either very tedious or even impossible.
Imagine you have recorded a complex interval session or fartlek and would like to recreate it as a structured workout. An LLM can do exactly that for you by analysing the recorded time series of your training session and creating a structured workout directly within a Tredict training plan, which you can then save as a template or load straight onto your sports watch for another attempt.
Or you can have an LLM analyse your training history, even across long periods of time, and get a summary of notable training sessions, see how your endurance performance has developed, or have the AI set better titles and descriptions for your training sessions, so that your training calendar becomes much tidier and you get a better overview of your training history directly in Tredict.
By analysing your training over the past months and examining individual strong sessions such as races, interval training or threshold runs in detail, an LLM can determine your capacity values, such as your FTP (Functional Threshold Power), FTPa (Functional Threshold Pace) or hrLth (Heart Rate Lactate Threshold). Combined with the automatic detection of these values in Tredict, this provides a particularly precise assessment of your current performance level.
And now the crowning feature: based on your training history and capacity values, an LLM can create a complete training plan directly in Tredict, including structured workouts that you can execute straight on your sports watch. The AI will typically even recognise whether you train based on power output, for example with a power meter on your road bike or a Stryd foot pod, or based on heart rate with a chest strap. Whether you are a triathlete, road cyclist, runner or swimmer, the AI creates the plan tailored to you directly within a Tredict training plan.
An important note is nevertheless warranted, as an LLM is no substitute for a coach! If you have never created your own training plans before, you should not adopt an AI-generated plan without reviewing it carefully. You may not be in a position to judge whether it truly suits you. Anyone who requires reliable expertise in endurance sports should seek out an experienced coach. AI does not change the relationship between athlete and coach, nor can it replace it, but it can significantly simplify existing processes for everyone involved.
Claude creates a reusable training plan for running and swimming directly in Tredict.
Significance for coaches and training professionals
Since coaches can create reusable training plans directly in Tredict with the help of an LLM, the tedious steps of manually creating individual segments of a structured workout, such as an interval session, are no longer necessary. As a coach, you simply describe in the AI chat what kind of structured workout you would like, and a reusable training plan containing that workout is created automatically, which you can apply directly to your athletes and which will then be loaded automatically onto their sports watch for execution.
Taking things a step further, a precise prompt to the LLM can generate complete training plans containing structured workouts spanning weeks and months! Using your professional coaching expertise, you then curate the created plans and build up, in less time, a library of reusable training plans and workouts that match your vision precisely while still bearing your personal touch. The time saved can in turn be invested in your personal work with your endurance athletes.
Yes, this really works! Give it a try.
What does not work? The LLM does not have access to your athletes' training data via the Tredict MCP server, as this data would otherwise be passed on to third parties. That is of course not a good idea.
Supported AI assistants and platforms
In general, any AI assistant is supported that acts as an MCP host or MCP client and can integrate an MCP server with authorisation. On some platforms this is referred to as an "MCP connector", while in agentic terminals it is often simply called an "MCP server".
Some platforms use the OAuth 2.1 authorisation method, while with others you can use a bearer token, which you can create under Settings → Personal API / MCP.
Ready-made connection guides
Here you will find a few ready-made guides for connecting the Tredict MCP server to popular AI assistants:
- Anthropic Claude Web, Claude Desktop
- Anthropic Claude Code
- OpenAI ChatGPT
- OpenAI Codex
- Mistral Le Chat
- Mistral Vibe
ChatGPT determines the athlete's current running FTP with the help of the Tredict MCP server.
Issues
LLMs work probabilistically rather than deterministically, meaning responses and training plans can vary with each request. Depending on the model and context size, sometimes quite significant differences can arise. Powerful LLMs with a large context window deliver remarkably good results even for complex training plans, while weaker models with a small context window or low limit quickly reach their boundaries with more complex structures and simply stop working. It is therefore partly up to you to find out which LLM you use best for which tasks in your training planning, especially as circumstances can change rather quickly. At present (as of February 2026), Claude holds the top position when used with the Tredict MCP server, consistently delivering excellent results for endurance training plan creation.When you use external AI platforms such as Claude by Anthropic or ChatGPT by OpenAI and connect the Tredict MCP server there, your training data will, depending on the permission settings in Tredict, be transmitted to these providers. Whether this is a concern is for each person to decide individually, though it does no harm to read the privacy policy of the respective platform. An alternative is offered by local or self-hosted models, for example Qwen with Ollama, which are already well suited today for simpler tasks such as improving activity titles or analysing training history, but quickly reach their limits with more complex structures such as training plans. As a European alternative to OpenAI and Anthropic, the models from Mistral AI in France have also made a name for themselves and can be used, including an MCP connector, directly on their own platform Le Chat.
Felix Gertz
Felix is the founder and developer of the endurance sports training platform Tredict. Since 2020, Tredict has been valued as a platform for training analysis and planning by endurance athletes and coaches around the world.