Hey everyone,
I'm a developer planning to build a cycling and running training app integrated with the Strava API, and before I start I'm trying to get clarity on the AI restrictions introduced in the November 2024 API Agreement update.
I understand that using Strava data to train or fine-tune AI models is clearly prohibited. What I'm less sure about is whether the restriction also covers inference-only use cases, specifically:
1. Sending a user's own Strava data as context in a single prompt to an external LLM API (Claude, GPT-4, Gemini etc.) to generate a personalized training plan or coaching feedback — no model training, data used only for that one request, result shown only to that user
2. Running a self-hosted local model on my own infrastructure (e.g. LLaMA, Mistral), where Strava data never leaves my server and zero training on Strava data occurs
Has anyone dealt with this and got a clear answer from Strava — either via the review process or by emailing developers@strava.com?
I've already sent an email to Strava asking for clarification, just curious if anyone here has been through this already.
Thanks!
