My current KICKR + ELEMNT companion app stack (and the AI piece I built)

Training
1

Saw the recent “Best Companion Apps for Analytics” thread and figured I’d share where I’ve actually landed after a couple years of bouncing between things. Maybe useful, maybe not :wink:

Stack right now:

  • SYSTM for the structured stuff on the KICKR, nothing’s beaten it for me
  • Strava as the dumb sync target - KICKR rides + ELEMNT outdoor rides land here
    and fan out
  • intervals.icu for CTL/ATL/TSB and the best-effort curves (free, does the job)
  • Oura for sleep context
  • An MCP server I built (that I call athletedata) so I can just ask ChatGPT or
    Claude questions across all of it

That last one - full disclosure, I made it. The reason I built it: I’d have SYSTM in one tab, Strava in another, intervals.icu in a third, Oura on my phone, and the question I actually wanted answered was always “ok given all this, what should today look like?” - which no one dashboard can really answer because they each only see one slice.

A few things that surprised me after using it for a while:

KICKR data is the most useful source by far. Wearable HRV/sleep numbers turned out noisier than the marketing made them sound. The boring “did I hit the watts on the trainer last week” data has been a way better predictor of how this week goes than anything my Oura tells me.

You don’t really need the LLM for most of it. Maybe 80% of the value is just “all the data in one place that something can query.” The chat part is the cheap layer on top - the unsexy part was building the integrations.

I haven’t gotten much out of ELEMNT outdoor ride data once I already have KICKR + Strava. Drafting messes with power, segments are gamed, GPS noise. I use the ELEMNT for navigation and that’s about it.

Curious if anyone here has actually pulled analytical signal out of outdoor rides that they couldn’t get from indoor + a wearable - because it kind of feels like a gap in my setup..

Comment/reach out if you want to try my MCP server to chat ith your data, it’s a 2min setup, you can try for free.

Curious to hear if you’ve built smth similar for yourself!

1 Like
2

I enjoyed SYSTM but wanted more free ride options so I moved to FulGaz and then Rouvy. I built a portfolio app that imports from Strava but uses GPS data for the mapping so SYSTM workouts won’t work. It does work with outdoor rides but I don’t use any sensors like power meter, cadence, or GPS bike computer, only HR and the Strava app.

Screenshot 2026-05-21 at 4.20.01 PM
Screenshot 2026-05-21 at 4.20.01 PM2992×1710 1.05 MB

Anyway to answer your question I do not get any thing more from the outdoor rides as there is less data, not more, but it does of course fill in gaps in the overall data.

The main benefit I’ve found is adding context, skills (data interpretation, coaching), and tools with AI - so an AI coach that understand me and adapts. I’m not training for events but like to keep fit and be competitive on Rouvy challenge rides (top 10-20 in my age group). I like flexibility. So rather than a rigid training program I can just set a goal like “ride undulating classics route comfortably at 4W/kg” and the AI coach builds and continuously adapts a program as it learns my patterns (for example I like doing a big climb on weekends). It can generate workouts (.zwo) for specific/favorite routes with commentary, that I just upload to Rouvy.

The AI coach does draw some lines that are “must do” to ensure tracking toward goal - I have the ATL,CTL,TSS, CP, W` metrics so easy to see, track and report on. It also uses these metrics and my feedback after rides (feeling, RPE) to when making suggestions - take it easier, type of ride to do today etc.

Overall what I like the most is the flexibility and that it understands me, my preferences, and adapts.

I also have an MCP server, main benefit with that is free-form chat/reporting that is not in the app UI. Here’s an example using Claude Desktop

Screenshot 2026-05-21 at 4.42.08 PM
Screenshot 2026-05-21 at 4.42.08 PM2148×1748 427 KB

Humorous feedback from the AI is fun: “The model is very good at telling you when you’re sandbagging the perception score, and this week it would gently raise an eyebrow at “GOOD, RPE 6” on a max-effort classification.”

As I said it’s a portfolio app that lets me learn/apply AI to my own data, not a commercial SaaS, too many cycling apps in the space already and key established main players in a small market: https://routebook.app

3

This is awesome! (just posting because Wahoo won’t let me keep posting in the threads I care about until I get my post count up by aparently posting in threads I don’t)

4

What extra information were you thinking of ? Fatigue effects or similar ?

5

Very cool + also very neat website, congrats!

6

Yeah fatigue + durability mostly - hour 3 power vs hour 1, decoupling under heat. Was hoping outdoor would show it cleanly but drafting/GPS noise gets in the way

1 Like