Connect Training Data to ChatGPT or Claude

You've tried it. Screenshot of last week's Garmin training load, paste into ChatGPT, ask "what should I do tomorrow?". The reply sounds smart. Then you remember it doesn't know your last 12 weeks, doesn't know you skipped Tuesday because your kid was sick, doesn't know your bench is stalling, and has no memory of yesterday's chat. You start over. Every. Single. Time.

This guide walks through what actually works when you want an LLM to coach your training, why the copy-paste workflow breaks for real athletes, and the cleaner pattern that's been hiding inside a purpose-built training app the whole time.

Why the copy-paste-to-ChatGPT workflow breaks

ChatGPT and Claude are excellent reasoning engines. They're not training databases. The moment you treat a general-purpose LLM as your coach, you take on three jobs the model can't do for you.

First, you become the data pipeline. Every conversation starts cold. You paste a screenshot of TSS, then a screenshot of HRV, then type out last night's sleep, then describe how the legs felt. Five minutes in, you've spent more energy formatting context than the model spends generating advice. Miss a screenshot and the advice is wrong, but the model won't tell you it's wrong, it'll just answer anyway.

Second, you become the memory layer. The model has no idea you ran a marathon eight weeks ago. It doesn't know your cycle phase, your masters age-group, your shoulder history, or that you train fasted three mornings a week. You can paste a long preamble every time, but most people don't, so the advice drifts toward generic.

Third, you become the corrector. LLMs hallucinate confidently. Without ground-truth data flowing in automatically, the model will happily prescribe a 90-minute zone 2 ride for someone who only has 45 minutes today, or stack a heavy squat session the day after a 20-mile long run. You catch it, or you don't.

The screenshot-to-ChatGPT loop isn't a coaching workflow. It's a riddle you give the LLM, and the answers are only as good as your riddle.

The actual fix: bring the LLM to your training data, not the other way around

The right architecture is the inverse of what most people try. Instead of exporting data into a chat window, you put the chat window inside the app that already has your data. The LLM gets structured, current, complete context on every turn, and you get to ask questions like "should I deload this week?" without typing a single number.

That's exactly what Movement Rebels does. The coach inside MR is the same class of model (Claude, Gemini, route-by-task) wrapped in a layer that already sees:

• Every Garmin workout for the last 12+ weeks via the native Garmin OAuth integration, including TSS, training load, HRV, body battery, and sleep
• Apple Health data via the native iOS app's HealthKit integration, which is also where WHOOP, Oura, Polar, and COROS data lands today
• Strava activities via the native read integration, with rich session summaries written back to Strava descriptions
• Your strength log from the in-app exercise codex, including PRs, baseline tests, and progression on every named lift
• Rebel Fuel: meal logs, snap-meal photos, macro estimates, hydration, supplement tracking
• Biohack history: breathwork sessions, NSDR, cold exposure, fasting timer, deadhang baselines
• Body comp trends and the long-form athlete profile (cycle phase, masters age, postpartum stage, sport-specific goals, injury history)

When you ask the coach "what should I do tomorrow?", it doesn't ask you to paste anything. It already saw you went to bed late, your HRV is down 14 percent, your Rebel Fuel log is 600 kcal under target three days running, and you have a planned tempo session. The reply is grounded in your actual week, not your description of your week.

What "training-native" buys you that general LLMs can't fake

The difference shows up in three places.

Memory across sessions. The MR coach holds an Athlete Memory that compiles from every conversation. Mention you tweaked your left knee in March, and six months later when you ask about your run plan, the coach still knows. ChatGPT forgets the moment the window closes.

Cross-domain reasoning. A real coaching question is rarely about one data stream. "Why am I bonking on long rides" is a question about Rebel Fuel intake, glycogen replenishment, sleep, training load distribution, and possibly the menstrual cycle phase. The MR coach pulls all of those at once. ChatGPT pulls what you pasted, and nothing else.

Action, not just advice. When the coach decides you need a deload, it doesn't just tell you, it adjusts the adaptive weekly plan and pushes tomorrow's session to your Garmin watch as a structured workout. When it recommends a recovery breathwork session, the breathwork timer is one tap away in the same app. When it spots that you're underfueling, it points you at Rebel Fuel and the snap-meal photo logger so the fix is one tap, not a separate app.

One honest note on Strava

If you're using Strava for the social side, the kudos, the friends' feed, the segments, keep using Strava. Movement Rebels does not have a social feed, and we won't pretend it does. MR replaces Strava's training value, planning, logging, AI coaching, recovery read, and fueling. The MR↔Strava integration is read and write, so your activity still posts to your friends' feed via Strava with a rich session summary the coach writes for you.

How to set this up

If you've been copy-pasting into ChatGPT or Claude, here's the cleaner version.

1. Sign up for Movement Rebels at app.movementrebels.com or download the native iOS app.
2. Connect Garmin (native OAuth, takes 30 seconds). If you're on WHOOP, Oura, Polar, or COROS, those flow through Apple Health on the iOS app, also one-tap.
3. Connect Strava if you want the read+write loop.
4. Spend 5 minutes in onboarding so the coach knows your goal, sport, history, and constraints.
5. Open the coach chat and ask the same question you'd have asked ChatGPT. Notice the answer references your actual data.

Pricing

Free 7-day trial, full access to the coach, plan generation, snap meal, and every tool in the app. No card required. After the trial, Pro+ is $20/month for unlimited coaching, adaptive planning, and all integrations.