Since 1 June 2026, Strava offers an official MCP connector with its Strava MCP server. That is a clear signal: the Model Context Protocol has reached the mainstream. Time for a comparison with the Tredict MCP server, which has been available since February 2026.
What the Strava MCP server can do The Strava MCP connector is a remote MCP server that gives Strava subscribers conversational access to their own activity data via Claude. In concrete terms, that means: activity history, fitness trends, GPS data, heart rate streams, cycling power data, as well as club and event data.
The important limitation here is that the connector is read-only. Data can be queried but not modified. On top of that, it is exclusive to Strava subscribers, currently only available through Claude, and the rollout is still being done in stages.
In the context of the API changes announced at the same time, the direction is clear, Strava is significantly restricting third-party access via the regular API and positioning its own MCP as a controlled, secured channel for data access by AI assistants. Whether this model is compatible with the GDPR right to data portability and the EU Data Act is an open question.
What the Tredict MCP server can do and what sets it apart
The Tredict MCP server covers the field of endurance sports training analysis and planning far more broadly, and in a GDPR-compliant manner. The decisive difference, however, lies in write operations.
Tredict can write, Strava cannot.
- Create reusable training plans
- Set up structured workouts and schedule them
- Move planned sessions to a new date
- Update activity titles and descriptions
These workouts are then automatically transferred to compatible platforms from Garmin, Coros, Zwift, Suunto, Wahoo and, via Watchletic, to the Apple Watch, provided the athlete adds them to their own training calendar. This closes the loop between AI assistant, training platform and device.
On the read side, Tredict additionally provides access to:
- HRV data and sleep values
- Body values (weight, resting heart rate, body fat, etc.)
- Capacity values such as FTP, FTPa, HRMax and lactate threshold
- Zone distributions and zone history
- Equipment list incl. shoe mileage and bike usage
- Training effort time series
On top of that, there are predefined prompts for common analysis tasks: year-on-year comparisons, in-depth analyses, FTP determination from training history, or creating a structured workout from an activity that has already been completed.
The Tredict MCP server also supports ChatGPT, Perplexity.ai and, with a personal API key, even any generic MCP client, not just Claude. And it is not tied to a subscription that primarily benefits the platform itself.
Those who wish to use ChatGPT also have a particularly straightforward way in. The Tredict ChatGPT App makes it possible to analyse training data directly in ChatGPT, and with a free ChatGPT account at that, no OpenAI subscription is required.
ChatGPT analyses the training history and creates a structured workout in a training plan directly in Tredict.
MCP is far from dead Strava, with its 195 million users, is one of the largest fitness platforms in the world. The fact that they are now releasing an MCP server is relevant for the entire industry. It will raise awareness of MCP as a standard. So much for "MCP is dead", as one heard rather often in the tech world.
For endurance athletes and coaches who want to seriously analyse and plan their training with AI, the Tredict MCP server remains the more comprehensive solution. If you want to do more than just read, if you want to plan in a structured way and transfer workouts directly to your watch, the Strava MCP server alone will not get you very far. The two approaches are not mutually exclusive. Anyone who tracks activities via Strava and uses Garmin or Coros can use the Tredict MCP server alongside it to make the most of AI-assisted training planning in endurance sports.