We are not taking your work away — we are taking the detail work off your hands. You decide the methodology, the model and the pricing; the system works out the daily level, sends it to the watch, and evaluates it every morning.
You have probably met an AI training planner before, and you were probably disappointed: it hands out a generic recipe, then leaves the athlete alone, and you have no role in it at all. Here the professional decision stays yours — the system fills in the frame that you draw.
This is the point where most AI tools fail in the eyes of a professional. Many systems have a single, pre-decided methodology behind them: always 80/20 polarised, or always a pyramidal distribution. Convenient for the developer — but it means the beginner preparing for a marathon gets the same thing as the seasoned racer sharpening for a competition.
| Setting | What it governs |
|---|---|
| Training methodology | Which methodology the system should follow — or leave it to pick one for the athlete's goal |
| Rest model | How it distributes recovery across the cycle — the rhythm of load and recovery weeks |
| Metabolic intensity | How aggressively it should lean on the measured metabolic profile |
| Phase emphasis | Which preparation phase to put the weight on |
You don't have to decide on all four to get started. Whatever you leave open, the system decides based on the athlete's goal.
Beyond the dropdowns you can also give your own free-text instructions, up to 2000 characters. Anything you would tell an assistant about the guiding principles of the preparation. All three channels can be switched on independently:
Plan instructions — applied automatically in every future generation.
Zone instructions — for the principles of zone calculation.
Feedback instructions — for the tone of the daily evaluation.
An important difference: the system always treats the athlete's text as data, never as an instruction. Your instructions, on the other hand, are real instructions — deliberately. This is your professional authority inside the system.
If you want something different for one particular athlete just once, you can override any of your settings when you start plan generation — for that one occasion only. It does not touch your saved defaults.
Plan generation and a one-paragraph daily evaluation are two tasks of completely different difficulty. They used to run on the same model — meaning that raising plan quality carried the same surcharge on the daily evaluation. Today you can choose a model separately for five tasks.
| Task | Why it is worth deciding separately |
|---|---|
| Training plan generation | The most complex task: this is where a higher-capability model pays off most. |
| Avatar conversation | Many short exchanges, low stakes — usually fine on a cheaper model. |
| Workout evaluation | Daily, short text. High volume, small context. |
| Weekly summary | This becomes the input for the next plan — worth spending more on than the daily one. |
| Lab analysis | Machine vision + numeric extraction. Rare, but accuracy-critical. |
If you set nothing for a task, it simply inherits your general model choice — there is no "automatic" option that would need separate handling. Model choice directly affects the AI cost you carry, and the range of available models grows along with the provider's offering.
The AI calls of the athletes you refer run on your own Anthropic API key. The platform does not bill you per token: you buy tokens directly from Anthropic, into your own account, and your athletes use those.
You see what you spent on Anthropic's own interface — no intermediate accounting, no markup on the token.
Per task. A model's price and capability move together, so this is a real cost/quality decision, and it is yours.
You set the price your athletes pay, within a multiplier band — the upper limit is tied to real model costs.
Until you set up your own API key, the athletes you refer cannot generate plans, talk to the avatar or have a lab report analysed. This is not a bug but deliberate: there is no silent fallback to the platform key, because that would push your athletes' costs onto the platform.
Setting up the key is a one-off, few-minute step — before saving, the system verifies with a live test call that it works. If a paid-for training plan is not created because the key is missing, that counts as a failed service delivery: the athlete gets an automatic refund, and the related commission is reversed as well.
This is the most common misunderstanding, so we say it sharply: the service package is chosen by the athlete at the time of purchase, and the content of the packages is uniform across the whole platform. You influence the methodology, the model and the price — not how much AI the athlete gets for their money.
| You decide | The athlete decides | The platform fixes |
|---|---|---|
| Training methodology, rest model, metabolic intensity, phase emphasis | Which package they buy: Basic, Premium or Ultra | What the three packages contain — uniformly, for every coach |
| Your free-text instructions on all three channels | When they buy, and whether they buy again after the block expires | The length of a training block: 4 weeks (28 days), with no grace period |
| Which AI model runs on the five sub-tasks | Which metric a session should run on (heart rate / pace / watts) | The base price and the upper limit of the multipliers |
| Your price multiplier within the given band | What data they disclose about themselves at all | The level of the platform fee — with your acceptance (see below) |
| Whether you start a plan generation for one of your athletes | Whether they follow the plan — the plan is advisory in nature | The data protection limits, at database level |
Coaches used to set the service level per athlete. That mechanism is gone: the screen and the endpoints behind it were both retired. Package selection moved to where the payment happens — to the athlete.
When an athlete you referred buys a training block, the amount arrives in your Stripe Connect account, and the platform deducts a 20% commission from it. The actual payout happens 15 days later.
They pick a package and pay — the amount set by your pricing.
The commission goes to your Stripe Connect balance automatically and immediately, with the platform fee deducted.
The purpose of the delay is that the athlete's 14-day right of withdrawal and any disputed transactions can still be settled.
You issue the invoice for your commission — that is why we ask for your tax number at registration.
| You handle | The platform handles |
|---|---|
| Setting the professional frame | The entire technical operation |
| Your own Anthropic tokens | The plan generation logic and its development |
| Keeping in touch with your athletes | The Intervals.icu integration and sync |
| Your own invoicing obligations | The payment infrastructure (Stripe, Billingo) |
| Data security and GDPR compliance |
The platform sets a base price. On top of that comes your price multiplier, whose upper limit is tied to real model costs — you have room to move, but pricing never detaches from the actual cost behind it.
On top of that you can also override the surcharge of the Premium and Ultra packages instead of the platform defaults (1.25× and 1.50× by default); the only rule is that no package can be cheaper than the base price, and the more expensive package cannot be cheaper than the cheaper one.
The level of the platform fee may change over time, but it does not automatically apply to an existing coach. If the platform raises the fee, you receive a proposal, and you have 21 days to decide whether you accept it. Until you accept, you carry on with your old, accepted fee.
A later fee change does not apply retroactively to purchases already completed — every purchase keeps the terms that applied at the time.
If the athlete paid but the plan was not created for technical reasons, the system automatically refunds the full amount and reverses the related commission as well.
If this happens within the 15-day delay, it runs through technically and you carry no further obligation. If the payout has already been made, the reversal may fail — in that case the platform offsets the amount against your next settlement.
Coach registration is self-service, but because you have to issue an invoice for your commission, the system automatically verifies your tax number when you save your billing details: a Hungarian tax number against the NAV register, another EU tax number through the European Commission's VIES system.
The billing details are saved, and your invite link becomes available — from that point on you can refer athletes.
The system rejects the save, nothing is written. Based on the error message you can correct the data and try again.
If the official system is temporarily unavailable, or the tax number was issued outside the EU, the data is saved but the invite link does not open until an administrator approves it manually. We decide on this within a reasonable time.
The condition applies to already registered coaches too: if your tax number is still in "unverified" state, your invite link is currently not visible. The fix is simple — save your billing details again, which runs the check. This restriction applies only to referring new athletes : it does not affect the athletes already assigned to you, the commission on their future purchases, or the commissions you have already earned.
The platform is built so that the athlete's most sensitive data never reaches a third party — including you. This is a data protection commitment made to the athlete, and it is enforced at database level, not in a house rule. You do see the range needed for professional follow-up: the plans, the content of the sessions and whether the athlete completed them.
This visibility is read access only: you cannot modify or delete the plan or the sessions — you influence the content of the plan by starting a new generation. You may use the data you learn solely to support that athlete professionally; the partner agreement imposes confidentiality with no time limit and purpose limitation.
Why is this good for you? Because your professional influence works through the frame, not through seeing data: your settings are present in every generated plan, even the ones you did not start. And because you don't have to go through every athlete's every daily data point — you would get stuck at a few dozen. The professional decision is your job, the data processing is the system's.
This is probably the question you care about most. Let's go through it.
| Step | Roughly how long |
|---|---|
| Coach registration + billing details, with tax number | 5–10 minutes, once |
| Obtaining and pasting in an Anthropic API key | 10–15 minutes, once |
| Connecting a Stripe Connect payout account | 10–20 minutes, once |
| Thinking through the methodology settings | 20–30 minutes, once |
| Model and pricing decisions | 10 minutes, once |
One afternoon is plenty, and after that you won't need to touch it for months.
| Runs automatically | Only if you want it |
|---|---|
| Plan generation according to your settings | Starting a plan for an athlete |
| Sync to the athlete's watch | Fine-tuning a setting |
| Daily evaluation and weekly summary | Changing a model or a price |
| Invoicing and commission payout | A one-off override on a generation |
| Tax number check on saving | Looking at an athlete's plan |
You don't have to deal with the daily "how much should they run on Tuesday" questions. You draw the professional frame once, and from then on the system works within it. You can spend your time on what is actually professional work — and with more athletes than you could otherwise handle.
It helps with platform usage and general sports-science questions. It does not receive or process athletes' health or training data.
With your own key: 30,000 tokens/day, at your cost.
Without your own key: 8,000 tokens/day at the platform's cost — but only if your tax number is verified or an administrator has approved your billing details. This exists precisely so that you can get an answer to the question "how do I set up my key".
A good choice if you work with several athletes or would like to, you have a settled professional view of preparation, and you would happily automate the daily detail work. It works particularly well with multi-sport athletes, or if you build on lab data.
Less of a fit if you work with one or two athletes in a very close, daily relationship and want to design every session personally. In that case the system would be extra work rather than less.
At sign-up you choose a coach account and give your tax number — this proves you can issue invoices.
→ the tax number check runsRegister with Anthropic, top up your balance, generate a key, paste it into the interface.
→ live test call before savingFor your payouts. It can be started once your billing conditions are in order.
→ 15-day payout scheduleFour dropdowns on the AI instructions page, plus free-text instructions if you want them.
→ affects every future generationPer-task model selection, price multiplier, and the Premium/Ultra surcharges if you want to override them.
→ changeable at any timeWith your own invite link. Anyone who registers through it is automatically assigned to you.
→ requires a verified tax numberBecause this way there is no intermediate accounting and no markup. Your athletes' AI calls run under your own Anthropic contract, you see the cost itemised on Anthropic's own interface, and you decide how much each model consumes. In return, setting up the key is your responsibility: without it, the AI features of the athletes assigned to you do not work.
No. This used to be possible, but it was retired. The service package (Basic / Premium / Ultra) is chosen by the athlete at the time of purchase, and the content of the packages is uniform across the whole platform. You influence the methodology, the model selection and the price.
The AI services of the athletes assigned to you fall back to the platform's own access, so that they are not left without service. This in itself does not affect your commission entitlement.
Four weeks (28 days), with no grace period. After that the athlete's AI-based services stop, but the finished plan, the zones, the lab results and the Intervals.icu sync all remain. Further commission is generated by the purchase of a new block.
Yes — you see the planned / completed / skipped status for every session, and you see the content of the sessions too. What you do not see: the actual performance data (heart-rate curve, watts, HRV, sleep), the lab values, the zones and the health data. Those are locked away at database level.
Yes, under a coach partner agreement — this governs the order of commission payouts, the cases of automatic refund, the tax number verification, and your confidentiality and purpose limitation obligations attached to plan visibility.
Open a coach account, set your axes, and see on one athlete what the system makes of it.
Open a coach account
for coaches