What Harlen is, the wearables it reads, how it compares to a dashboard, and how personal trainers put it to work, in plain English.
Harlen is an AI assistant. It holds a context model for every client on your roster, reads their data against their own baseline every night, and does the work: the scan, the block, the check-in triage, the calendar. Nothing reaches a client until you send it.
Harlen is built for personal trainers who run a roster of clients and already track wearable data. It is most useful for trainers managing more clients than they can hold in their head, who want to keep individual attention without spending hours reading raw data by hand.
Harlen scans a coach's whole roster before the first session of the day and surfaces the clients who have drifted from their own normal. For each one it explains what changed, the likely reason, and what to adjust in the next session, with the data sources cited. The coach walks in already knowing where to look.
Restoration is Harlen's recovery readiness score for a client, measured against that client's own history, so the reference point is who they usually are. When it dips below a client's baseline, Harlen flags it so the coach can adjust volume before the client trains tired. We use the term Restoration to keep it distinct from any single wearable's branded readiness metric.
Garmin, Polar, Withings, MyFitnessPal, and Apple Watch (through Apple Health) are all active today. Whoop works through Apple Health, though it does not pass HRV or detailed sleep metrics. Coros, Oura, and Fitbit are in beta while we test them. Harlen also reads the lifestyle and training data a coach logs directly, and if a client wears a device that writes to Apple Health, Harlen can generally read it. The current list is on the docs page.
Apple Watch and Garmin are active today, with Apple Watch connecting through Apple Health. Oura is in beta while we finish testing it. From each of these Harlen pulls recovery, sleep, heart rate, training load, and activity data, and reads every metric against that client's own baseline.
Harlen combines four kinds of input: wearable data (recovery, sleep, heart rate, training load), lifestyle context (stress, travel, nutrition), what the coach logs (notes, check-ins, programming), and the client's own history. It connects these into one model per client, so you read one picture instead of showing them as separate dashboards.
Yes. Client data is used only to power the coaching reads inside a coach's own account and is never sold. Harlen is in beta and we are happy to walk through our data handling in detail on a demo call.
A dashboard shows you data and leaves the interpretation to you. Harlen makes the read: you get the conclusion first, with the supporting data underneath, so you can see who needs attention today without scanning charts to work it out.
ChatGPT is a generalist. It starts from nothing every time, so you would paste in the injury history, the last block and this week's numbers before it could say anything useful, then do it again tomorrow for the next client. It also cannot see the data: it has no connection to a watch or a ring, so every number reaches it only because you typed it in. Harlen is built for this one job. It holds a context model for every client on your roster, keeps it current, and reads their wearables against their own baseline overnight whether or not you open it. It also learns how you work, because every edit you make to a block teaches it and the notes you log become durable facts about that client. A generalist can be told your standards once in a prompt, and it forgets them by the next conversation.
Harlen uses large language models the way most modern software does, and the part that matters is what sits around them: a context model per client, a connection to their wearables, and a read tied back to their own baseline and to peer-reviewed research. Asking questions in plain language is one part of the system rather than the whole of it.
No. Harlen does the reading so the coach can make the call. It surfaces what changed and suggests adjustments, and the coach decides what to do. The goal is to carry a coach's judgment across a bigger roster.
Harlen uses AI to connect each client's signals into one model, detect meaningful change against their baseline, and explain it in plain language. Every read is tied back to the underlying data and to peer-reviewed research, so you can trace any output back to what produced it.
Yes. You can ask Harlen about any client the way you'd ask an assistant who has watched them for months, and get a synthesized answer grounded in that client's real data with the sources shown. There is no query syntax to learn.
Yes. Harlen can draft programming informed by a client's recovery, training load, and history, which the coach then edits and approves. It is built to fit how coaches already plan blocks and sessions, and every draft waits on your edit.
No. Harlen surfaces patterns in training and recovery data to support coaching decisions; it is not a medical device and does not provide medical advice or diagnosis. Coaches should refer clients to qualified medical professionals for health concerns.
Harlen is in beta, and access starts with a short founder-led demo. Pricing is not published yet and is discussed on that call once we understand your roster. You can book a demo from any page on this site.