This toolset creates a “curated” dataset out of your data in intervals.icu (rides only) and has everything to work on this with your preferred AI coach.
Option 1: “Bit and bytes” - run the provided Pythons scripts to create the files for GenAI Tools
Option 2: “Managed” Use the MCP server at https://intervals-mcp.training-architect.com to integrate this directly with GenAI tools (step-by-step guide for Claude, ChatGPT, Copilot provided)
Hi Robert, really cool project!
Is it targeting only rides by design or are you planning to extend it to other disciplines (e.g. running, weight lifting sessions, swims)?
Looks cool!
There is an existing Intervals - MCP Server project called IcuSync https://icusync.icu/ - just FYI, in case it causes confusion.
Thx for your feedback. Next step in the roadmap is to provide this as a web application. After this I am open for more - e.g. supporting other endurance activities. We will see …
Yes, I have seen this one. The difference: My project is “open book”, that means you can see what is done, and how it is done. For free … And you have a “coaching system” with system prompt, coach logic etc.
Nice! I just meant the similarity in the names …good work!
Popping my head in to say I’ve been using this for the last week or so and think it’s really good. Excellent work.
@Robert_Brands I love it being open source! I will have a deeper look at it but it looks promising and already quite complex. Do you use it personally running the scripts directly or do you have some unreleased CLI tooling to make your life easier?
@lucasp I am exactly using this as it is - no unreleased tools! I have my environment with Visual Studio Code, in the terminal window I run “python .\scripts\prepare_week_for_coach.py” and after getting a plan for the week from ChatGPT & Co I run “python .\scripts\upload_plan.py”. That’s all. But now as the MCP server is available I use it directly inside of the GenAI tools.
The latest updates https://github.com/rbrands/intervals-icu-sync/blob/main/CHANGELOG.md:
See intervals-icu-sync/CHANGELOG.md at main · rbrands/intervals-icu-sync · GitHub for the latest updates:
list_library_workouts and list_standard_library_workouts This allows access to a set of standard workouts from your library or a centrally managed library (from me).See intervals-icu-sync/CHANGELOG.md at main · rbrands/intervals-icu-sync for the latest release:
For me all this is working quit well the last weeks. Let me know how it works for you.
See intervals-icu-sync/docs/gen_ai_setup_step_by_step.md at main · rbrands/intervals-icu-sync how to start.
See intervals-icu-sync/CHANGELOG.md at main · rbrands/intervals-icu-sync · GitHub for the latest release. The dataset was optimized to minimize token usage. To use it reload input_schema.md from coach_logic.
See intervals-icu-sync/docs/gen_ai_setup_step_by_step.md at main · rbrands/intervals-icu-sync · GitHub for how to set up your environment.
See intervals-icu-sync/docs/gen_ai_setup_step_by_step.md at main · rbrands/intervals-icu-sync · GitHub for another update. Further optimization of the dataset. And now categorization of FTP, VO2Max and rider type included.
Option 3: “Integrated Web App” https://training-architect.com is now available. It runs the whole workflow — fetching your data, assessing it, generating a plan and uploading it — end to end in a single interface. There’s no local setup and no manual file exchange; the analysis, the plan and the calendar all live in one place. It’s for riders who’d rather not assemble the pieces themselves and want the entire loop under one roof.
See Three Tiers, One Method · Training Architect for an overview on the three options.
The idea of intervals-icu-sync is a “curated dataset” for easy handling in GenAI tools. Now, there is an option to get it with tools like Postman etc.. Use training-architect.com/api/dataset. https://training-architect.com/swagger defines the interface.
Everything about the “curated dataset”: The Curated Dataset — What the Coach Actually Sees · Training Architect
New version: Cleaning up some datapoints. intervals-icu-sync/CHANGELOG.md at main · rbrands/intervals-icu-sync · GitHub
Thank you for packaging this up and sharing it. I will have to give this a try.
It’s very educational to see the architectural decisions. It’s interesting shipping the coaching logic as a skill, I like the separation of the data vs the opinion.
As a user of the other ICUSync, I think I am going to have to rename this to get Claude to disambiguate between the two connections.
I have ChatGPT Plus. The guide says to add the MCP server, but I don’t have “Connectors” or “Add custom connector” in ChatGPT. Is this feature currently limited to specific ChatGPT plans or staged rollout?
Hi Robert,
I have ChatGPT Plus and I’m following your gen_ai_setup_step_by_step.md guide. However, my ChatGPT account only shows Apps/Plugins and does not have “Connectors”, “Developer Mode” or “Add custom connector”.
Is direct MCP integration with ChatGPT currently available for standard Plus accounts, or is it limited to specific ChatGPT plans or staged rollout?