Reps turns Apple Watch recovery data into an automatic, day-by-day training editor

Table of Contents

1. Key Highlights
2. Introduction
3. How Reps turns a 0–100 recovery score into a training decision
4. Inside the readiness score: inputs, weights and calibration
5. HRV: the measurement problem and Reps’ compromise
6. From score to session: the decision engine and ceilings
7. A worked example: the same Tuesday at 52 and at 88
8. Where the loop stops: user control, safety and AI
9. Data, integrations and the Apple Watch requirement
10. Pricing, platform requirements and Pro features
11. How Reps compares with WHOOP, Garmin, Oura and other adaptive training apps
12. Practical guidance: how athletes should use Reps
13. Potential pitfalls and areas for improvement
14. The developer, release cadence and adoption
15. Who should consider Reps and who should not
16. Real-world examples that illustrate value and limits
17. What to watch for next
18. FAQ

Key Highlights

• Reps reads Apple Health and Apple Watch signals each morning, computes a 0–100 readiness score, then applies an intensity ceiling that edits the day’s planned workout before it’s generated.
• The readiness composite weights sleep, subjective wellness, HRV, and training load heavily; Reps uniquely combines last-night HRV with a seven-night trend and redistributes missing inputs instead of penalising them.
• The app requires iPhone and Apple Watch to enable its closed-loop adjustment; it writes constrained workouts to your plan and offers an AI coach, but safety-critical alerts and explicit user requests override automatic changes.

Introduction

Athletes already live with numbers: Body Battery, Training Readiness, Oura’s sleep score, WHOOP’s recovery percentage. Those metrics sit on a screen one swipe away from the training plan, leaving a human to decide whether to keep the session, tone it down, or cancel. That choice happens groggy, quickly, and without clear rules. Reps reframes the final step as the product itself. Rather than delivering another 0–100 score, the app converts physiologic and behavioural signals into a concrete training decision and edits the day’s session accordingly—before the athlete ever taps Start.

Reps is the work of a single developer from Toronto, Siddharth Natamai. Designed for iPhone and Apple Watch users, it pulls data from Apple Health, runs a multi-layer decision engine, and returns either a constrained workout or an unchanged session depending on the athlete’s readiness. The premise is simple and pragmatic: measurement without actionable change is incomplete. For athletes who want reliable, reproducible daily guidance rather than a number to interpret, Reps proposes a different workflow. The design choices behind that workflow determine both its usefulness and its limits. The remainder of this article breaks down how the system operates, what goes into its verdicts, real-world scenarios that show its effects, and what athletes should know before integrating it into training.

How Reps turns a 0–100 recovery score into a training decision

Most wearables and apps stop at a single readiness number and leave the rest to the athlete. Reps continues through three additional stages: it grades recovery domains, derives a maximum-intensity ceiling, and then rewrites today’s planned entry in the weekly schedule. Finally, that ceiling travels with the workout-generation request so the training session created is already constrained.

The pipeline runs like this each morning:

1. Score: Reps ingests Apple Health data and computes a composite readiness number with one of five bands—Optimal (85–100), Good (70–84), Moderate (55–69), Low (40–54), Poor (0–39).
2. Decide: Four domains—autonomic recovery, sleep quality, training-load safety, and injury risk—are graded Green, Amber, or Red. The worst grade produces an intensity ceiling of Unrestricted, Moderate, Light, or Rest Only.
3. Rewrite: The ceiling edits the workout scheduled for that day. High-intensity sessions become moderate or light; Rest Only replaces the session with an active recovery alternative.
4. Generate: The constrained parameters are included in the request to the workout generator and the AI coach, so the session that arrives is already appropriate for the athlete’s current state.

The crucial distinction is practical: Reps does not merely advise “train conservatively”—it proposes a concrete change with a reason attached. Athletes see the suggested modification and can accept or decline. Explicit user requests override the recommendation, while safety-critical alerts (for example, an irregular rhythm flagged by Apple Watch ECG) cannot be dismissed. That balance aims to marry automated guidance with athlete agency.

Inside the readiness score: inputs, weights and calibration

Reps’ readiness composite is not a mystery. The app intentionally exposes the pieces and their weights so athletes can see why a number looks the way it does.

The headline allocations:

• Sleep: 19.4%
• Subjective wellness (user self-report): 18.9%
• Heart-rate variability (HRV): 15%
• Training load: 12.3%
• Previous-day protein intake: 7.6%
• Resting heart rate: 7.2%
• Hydration: 6.6%
• Alcohol (previous 24 hours): 5.7%
• Respiratory rate deviation: 4.7%
• Walking symmetry and steadiness: 2.8%

Two design principles stand out. First, sleep and subjective feeling together account for nearly 40% of the score, reflecting that both objective staging and how an athlete perceives recovery matter for daily readiness. Second, HRV receives a substantial share but is handled carefully because Apple Watch sampling differs from chest-strap overnight measurements.

Missing inputs are not punished. If an athlete never logs alcohol, Reps redistributes that weight across the available signals rather than zeroing it out. That prevents silent penalisation for absent data and keeps the score meaningful from day one. When insufficient historical data exists to form a personal baseline, the app labels the score as calibrating rather than presenting it as a definitive verdict.

Real-world implication: on a travel-heavy week with poor sleep and irregular food logs, Reps will still function and explain its uncertainty rather than deliver a false sense of precision.

HRV: the measurement problem and Reps’ compromise

HRV has become a sine qua non of readiness calculations, but not all HRV is the same. Chest straps and dedicated monitors typically produce a clean overnight SDNN or RMSSD baseline. Apple Watch, by contrast, samples opportunistically across the day. Early versions of Reps excluded HRV entirely because a single afternoon reading could appear far worse than the true overnight baseline.

The solution is statistical: Reps now extracts the median SDNN from qualifying nights (nights with at least three hours of actual sleep, excluding naps). The 15% HRV weight divides into two components: 8% for last night and 7% for the seven-night rolling trend. Each component is compared against the athlete’s own historical baseline rather than a population average.

Why both? The rolling trend catches chronic maladaptation—ramps in training load that accumulate over weeks. The last-night component answers the immediate question: should I push today? A seven-night average lags acute changes; a single-night value can be noisy. Using both captures the necessary temporal scales without overreacting or being blind to an acute poor night.

Athletes who track HRV with a chest strap will still get a fuller picture if the strap writes into Apple Health. Those who rely only on non-Apple devices may see degraded HRV input because many vendors do not write HRV or resting heart rate to Apple Health.

From score to session: the decision engine and ceilings

Scoring is only the start. Reps assigns domain grades—Green, Amber, Red—across autonomic recovery, sleep, training load, and injury risk. The worst domain grade determines the ceiling:

• Unrestricted: no restriction; train as planned.
• Moderate: limit high-intensity elements; keep a meaningful portion of the session.
• Light: reduce intensity significantly; preserve volume or convert to low-load alternatives.
• Rest Only: replace the session with an active recovery option or rest.

The app shows the reason for the ceiling. For example, a Red sleep domain produces a card that might say “Poor sleep is severely limiting recovery. Prioritise rest and recovery today.” The rewrite then changes the scheduled session, tags it with the reason—e.g., “Intensity reduced — CDS ceiling is Light”—and presents it for the athlete to accept or override.

Two nuanced behaviors protect both training adaptation and athlete safety:

1. Training load is treated as an acute:chronic workload ratio using conventional bands. Below 0.8 signals detrainment; 0.8–1.3 is optimal; 1.3–1.5 invites caution; above 1.5 elevates risk. The ratio is calculated from Apple Health workouts and active energy.
2. Rest Only is not thrown casually. It requires specific triggers such as an irregular rhythm notification, a critical blood-oxygen measurement, or an explicit rest mode set by the user.

The system avoids a fully closed loop that could undermine coach-athlete relationships. An athlete can override most constraints. Safety alerts are the exception.

A worked example: the same Tuesday at 52 and at 88

Concrete scenarios illuminate the practical effect. Suppose an athlete has Tuesday scheduled as a 55-minute high-intensity upper-body session, written into the weekly plan on Sunday.

Scenario A: 7:00 a.m., readiness 52 (Low)

• Sleep quality: 47% (below the 50% threshold that flags sleep domain Red).
• Resting heart rate: elevated against a three-day rolling average.
• HRV: approximately half a standard deviation below baseline.

Two domains flip Red: sleep and autonomic recovery. The worst grade produces a Light ceiling. Reps rewrites the session: intensity drops to moderate, duration remains 55 minutes, and the card explains the reason. The generation request carries the Light ceiling, so the workout the athlete opens is a lighter session, not a truncated version of the original.

Scenario B: 7:00 a.m., readiness 88 (Optimal)

• All four domains Green.
• Ceiling: Unrestricted.

No changes are proposed. The session arrives precisely as planned and the generation request is tagged “No restrictions — train at any intensity.”

Between those poles, Amber in autonomic, training-load or injury domains caps the day at Moderate. A Red sleep domain alone produces a Moderate cap rather than Light in some situations, because Reps balances domains against each other rather than letting one poor metric dominate. If the ceiling drops to Rest Only, the session is replaced entirely.

This example highlights the practical difference between a score that tells you to “think about scaling” and a score that arrives with an actual, editable workout. The athlete receives a reasoned proposal with the option to accept, decline, or request a specific session.

Where the loop stops: user control, safety and AI

A system that rewrites workouts must define boundaries. Three limits matter:

1. Athlete agency: An explicit user request for a specific session outranks the readiness recommendation. Most constraints are individually overridable. That preserves autonomy and coach-athlete agreements.
2. Safety-critical alerts: Irregular rhythm notifications, critical blood-oxygen levels, or other Apple Watch medical-grade alerts produce non-overridable restrictions. Reps contextualises ECGs and other signals but does not diagnose; it will not reclassify a Watch ECG reading.
3. AI coach operation: Reps includes an AI coach, Rex. Most queries are processed in the cloud; lighter interactions may use Apple’s on-device models. The app labels whether a response came from the device or the cloud. Privacy-conscious athletes should read that label and understand the trade-offs between richer, cloud-based context and local processing.

Those limits are deliberate. A completely closed loop that automates everything would risk training errors and upset athletes who want to keep manual control. The curated override model keeps automation practical and accountable.

Data, integrations and the Apple Watch requirement

Reps operates exclusively through Apple Health for inbound data. There is no direct Garmin Connect, WHOOP, or Oura API integration. The only vendor-facing link is outbound and optional: Strava can receive finished Reps workouts but the app imports nothing from Strava.

Consequences for athletes using non-Apple wearables:

• What usually arrives via Apple Health: workouts, steps, active energy, and body weight, because many vendors write those.
• What does not: HRV, resting heart rate, stress scores, and vendor-specific metrics like Body Battery. Many competitors do not write those to Apple Health.

The practical implication is clear: Reps needs an Apple Watch to deliver meaningful readiness-based training adjustments. Without an Apple Watch there is no overnight sleep staging, no dense resting-HR series, and no Apple-based HRV sampling. The strength-logging and nutrition features still work on iPhone alone, but the closed-loop recovery-to-training adjustment requires watch-derived signals.

Inside Apple Health, Reps reads sleep stages, heart rate, resting heart rate, HRV, respiratory rate, blood oxygen, ECG results, VO2 max, workouts and routes, active energy, steps, walking steadiness and asymmetry, running dynamics, body composition, and nutrition. The app writes back workouts and GPS routes, logged nutrition and hydration, and body measurements.

Some connectors in the app are present but inactive. Garmin and WHOOP rows exist in settings but are inert because Garmin’s developer program is closed to new applications; Fitbit has a “Coming Soon” note. Athletes should treat those as absent.

Real-world example: a triathlete who runs with Garmin but also wears an Apple Watch for sleep will get a mostly complete Reps experience if Garmin steps and workouts sync to Apple Health. If the athlete expects Reps to use Garmin Body Battery or WHOOP HRV directly, they will be disappointed. The system depends on the Watch’s full dataset.

Pricing, platform requirements and Pro features

Reps requires iPhone on iOS 26 or later and an Apple Watch on watchOS 26 for the full skill set. There is iPad support, Home and Lock Screen widgets, Live Activities, and Siri Shortcuts. The readiness score itself, strength logging, manual nutrition entry and basic tracking are free.

Reps Pro unlocks the AI coach (Rex), generated workouts and plans, photo-based nutrition analysis, a caffeine model that recommends a daily cut-off time based on a roughly five-hour half-life and the athlete’s circadian profile, and a body-battery-style energy metric. Pricing at the US storefront is $9.99/month, $69.99/year, or a $99.99 lifetime purchase. Both subscriptions include a one-week free trial; regional pricing varies.

The caffeine model deserves mention because it demonstrates the app’s intent to act on causal factors. By modelling a five-hour half-life adjusted for the athlete’s circadian timing, Reps suggests when to stop caffeine to avoid degrading sleep and subsequent recovery. That directly ties an input (late-day caffeine) to an actionable recommendation that can alter tomorrow’s readiness.

For athletes who want coaching and automated session generation, the Pro tier is the point at which the closed loop becomes operational end-to-end. Free users still receive the readiness number and can manually adjust plans, but they forgo the AI-generated constrained sessions.

How Reps compares with WHOOP, Garmin, Oura and other adaptive training apps

The wearables and apps ecosystem fragments into measurement, advice, and automated adaptation.

Measurement: WHOOP, Oura, Garmin and Apple Watch provide differing quality and completeness of metrics. WHOOP emphasises recovery percentage and strain. Oura focuses on sleep staging and readiness. Garmin provides Body Battery and Training Readiness signals on higher-end watches. Apple’s Health ecosystem aggregates many raw signals in a privacy-forward way. Reps builds on Apple Health rather than replacing it.

Advice: Most vendors offer recommendations—train easy, recover, or push—but leave the execution to the athlete. Reps converts advice into edits to the workout schedule. That operational step is where Reps diverges most sharply.

Adaptive training apps: Apps like IntervalCoach also adapt workouts to recovery signals; IntervalCoach uses Intervals.icu data rather than Apple Health and integrates with different toolchains. Vitara emphasises data transparency and local processing. Reps' distinctive combination is the morning decision-to-edit flow built around Apple Health inputs and a constrained generation request.

Strengths of Reps compared with others:

• Actionable edits arrive with reasons and remain attached to the workout generation.
• HRV handling uses both last-night and rolling trend comparisons against personal baselines.
• Missing data is redistributed rather than zeroed, preventing silent penalisation.
• The user retains override control except in safety-critical situations.

Limits relative to competing ecosystems:

• Functionally dependent on Apple Watch; athletes invested only in Garmin or WHOOP hardware lose the recovery half of the product.
• No direct APIs for some major wearable vendors.
• The AI coach relies on cloud processing for some queries, which may trouble users concerned about off-device data processing.

Practical guidance: how athletes should use Reps

Reps will suit athletes who want morning clarity and prefer decisions that arrive ready-made. To use it effectively, consider these practical approaches.

1. Equip yourself correctly Use an Apple Watch for overnight sleep staging, HRV sampling, resting heart rate and pulse-ox. Without it, Reps is a logging and coaching app rather than a closed-loop training assistant.
2. Build baseline data The score’s calibration improves with history. Log consistent sleep, nutrition, hydration, and subjective wellness for at least a couple of weeks to let personal baselines stabilise. Reps flags calibrating states while it learns.
3. Use the proposed changes, but keep a manual override Treat Reps’ edited sessions as recommendations backed by data. If you’re preparing for a race and a planned key session aligns with a taper or an essential adaptation, use the override but do so consciously. Reps labels when a recommendation has been overridden so you can review risky patterns.
4. Respect safety-critical alerts Do not dismiss irregular rhythm notifications or critical SpO2 readings. Reps enforces non-overridable restrictions in those cases for legitimate medical reasons.
5. Combine with coaching plans thoughtfully If you have a coach or follow a rigid plan, discuss how Reps will interact with that plan. The app preserves explicit user requests—use that feature to lock in must-do sessions while allowing automated adjustment on routine days.
6. Leverage the caffeine model and nutrition analysis Use Pro features to connect behaviours to readiness. If late caffeine consistently pushes your score down, the cut-off time will help you optimise both sleep and training windows. Photo-based macro logging reduces friction and increases the accuracy of nutrition-related inputs.

Real-world scenario: a marathoner who trains with a coach might set race-specific intensity days as explicit requests so Reps does not reduce them during the taper. For other midweek sessions, the athlete can allow Reps’ edits to protect against chronic overreach.

Potential pitfalls and areas for improvement

No tool is perfect. Reps introduces automation but also creates new decisions and dependencies.

Data dependency and vendor lock Athletes who prefer Garmin, WHOOP, or Oura as primary hardware will find Reps’ dependency on Apple Watch limiting. A full-featured Reps experience requires the Watch. If Garmin or WHOOP eventually open APIs to allow richer inbound data, Reps could broaden compatibility; until then, users must plan hardware accordingly.

Overreliance on automation Automation can habituate athletes to accept suggestions without scrutiny. Coaches and athletes should periodically audit overrides and examine when and why Reps proposed reductions. If an athlete consistently overrides the app, the inputs or thresholds may require recalibration.

Edge cases in HRV and situational noise The median-night HRV approach mitigates single-sample noise, but weird nights—airline travel, illness onset, or medication—can still distort short-term signals. Reps’ redistributive handling of missing inputs helps, but athletes should also flag anomalous days explicitly in subjective wellness inputs.

AI coach transparency and privacy Cloud-assisted coaching increases capability but raises privacy and latency considerations. Reps labels when a response used cloud processing, but athletes should weigh the benefits against the desire for entirely local processing.

False alarms and medical interpretation Reps contextualises Apple Watch ECG and irregular rhythm notifications but stops at interpretation. Athletes should not use Reps as a diagnostic tool. Instead, they should seek medical review for clinically significant alerts.

The developer, release cadence and adoption

Reps launched on 10 October 2025 and has seen rapid iteration: 28 released versions in less than a year from a solo developer. The App Store review picture is small but positive—currently 4.3 stars from 18 ratings, 13 of them five-star. Rapid updates are both a strength and a signal that the product remains in active refinement; early adopters should expect frequent feature changes and tuning.

The single-developer model shapes both advantages and constraints. A focused vision allows coherent design choices and fast iteration. Resource limits explain why some vendor integrations remain inert and why certain features depend tightly on Apple platform capabilities.

Who should consider Reps and who should not

Consider Reps if:

• You own an Apple Watch and want your morning readiness number to translate into a concrete workout change.
• You prefer decisions delivered with reasons and the ability to accept or override.
• You want an AI coach and generated workouts that respect your current physiological state.
• You value a privacy-focused local data feed (Apple Health) but accept cloud processing for richer AI interactions.

Reps is not a fit if:

• You rely exclusively on non-Apple wearables and expect a full-featured integration.
• You need a medical-grade diagnostic tool.
• You prefer a fully manual coach or a system that never overrides planned sessions automatically.

Real-world examples that illustrate value and limits

Example 1: A busy runner with erratic sleep A runner who works shifts and loses sleep sporadically benefits from Reps because it turns an unpredictable night into a specific change: replace a tempo session with a low-impact technique session and preserve adaptation through consistent, lower-risk volume.

Example 2: A cyclist doing a high-volume block Cyclists who log long rides on a head unit that syncs to Apple Health will find the training-load safety domain useful. An acute:chronic workload ratio that creeps above 1.5 will generate Moderate or Light ceilings, preventing sudden overreach that often leads to injury.

Example 3: A strength athlete with no Apple Watch Strength logging and the AI coach still work on the phone. Reps will not rewrite sessions based on overnight HRV or sleep staging, so the athlete gains coaching benefits without closed-loop recovery adjustment.

Example 4: A travel-heavy triathlete Travel disrupts sleep and circadian rhythm. Reps’ caffeine cut-off model and respiration/SpO2 awareness help triathletes stabilise sleep windows and reduce the risk of starting key sessions while maladapted.

What to watch for next

Reps’ current constraints are primarily platform and API availability. Watch for:

• Broader vendor integrations if Garmin, WHOOP, or Fitbit open their APIs to provide HRV and resting-HR data into Apple Health or directly to Reps.
• More on-device AI features as Apple’s local models expand, which could shift more queries off the cloud.
• Community-derived presets and coach integrations that allow trainers to codify when to accept or override automated ceilings for specific training phases.
• Research partnerships that might refine the weights or validate domain trigger thresholds against injury and performance outcomes in diverse athlete populations.

A mature Reps would combine richer vendor data, configurable coach rules, and more local inference while preserving the app’s core promise: a morning verdict that turns measurement into an actionable day’s session.

FAQ

Q: Will Reps work with Garmin, WHOOP or Oura? A: Only indirectly through Apple Health. Workouts, steps, active energy and body-weight entries often sync from those vendors into Apple Health and will be available to Reps. HRV, resting heart rate, stress scores and vendor-specific metrics like Body Battery generally do not get written to Apple Health, so Reps cannot see them. There is no direct API-level integration with Garmin, WHOOP or Oura currently.

Q: Is an Apple Watch required? A: For Reps’ readiness score and the automated training-adjustment loop, yes. The app relies on Apple Watch for overnight sleep staging, dense resting heart-rate sampling, HRV, respiratory rate and certain medical alerts. Strength logging, nutrition features and the AI coach can operate without a Watch, but the closed-loop recovery-to-training model requires it.

Q: Does Reps actually change my workout or only advise? A: Reps applies an intensity ceiling to today’s entry in the weekly plan and includes that ceiling in the workout-generation request. The workout you open will be constrained already. The app presents the proposed change with a reason and allows you to accept or decline. An explicit user request for a specific session overrides the recommendation. Safety-critical alerts are non-overridable.

Q: How does Reps handle HRV given Apple Watch sampling differences? A: Reps now extracts the median SDNN from samples recorded during qualifying nights (three or more hours of sleep). It divides HRV’s 15% weight into an 8% last-night measure and a 7% seven-night rolling trend, each compared against the athlete’s personal baseline. That approach captures acute and chronic signals without overreacting to single-sample noise.

Q: What inputs matter most to the readiness score? A: Sleep (19.4%) and subjective wellness (18.9%) are the largest single contributors, followed by HRV (15%) and training load (12.3%). Other factors include previous-day protein, resting heart rate, hydration, alcohol, respiratory rate deviation and walking symmetry.

Q: What happens if data is missing? A: Missing inputs are redistributed across the remaining signals rather than being set to zero. If the app lacks enough history to establish a personal baseline, it labels the score as calibrating rather than presenting a definitive result.

Q: Can I override Reps’ recommendations? A: Yes. Athlete-requested sessions outrank automated recommendations. Most constraints are overridable, but safety-critical restrictions—such as irregular rhythm or critical blood-oxygen notifications—cannot be dismissed.

Q: Is Reps a medical device? A: No. Reps contextualises Apple Watch medical alerts but does not diagnose or reinterpret ECG classifications. It provides training-context advice, not clinical assessment.

Q: How much does Reps cost? A: The app is free for tracking, readiness scoring, strength logging and manual nutrition. Reps Pro—unlocking the AI coach, generated workouts and plans, photo-based nutrition analysis, the caffeine model and a body-battery metric—costs $9.99/month or $69.99/year in the US, with a $99.99 lifetime purchase option. Prices differ by region and include a one-week free trial.

Q: Where can I download Reps? A: Reps is available on the App Store and at reps.siddharthnatamai.com.

Q: Who built Reps and how mature is it? A: Reps is developed by Siddharth Natamai, a solo developer based in Toronto. It launched on 10 October 2025 and has seen frequent updates—28 released versions in its first months—indicating active development and iteration.

Q: Will Reps replace my coach or training plan? A: Reps is a tool to augment decision-making. It rewrites sessions based on readiness signals but preserves athlete agency and coach directives when a session is explicitly requested. Treat Reps as an assistant that highlights when your body needs a different stimulus, not as a wholesale substitute for a trained coach’s periodised plan.

Q: What should I monitor once I start using Reps? A: Track how often you accept or override recommendations, whether the app’s ceilings align with your subjective feeling, and whether injury risk or performance trends shift over weeks. Use the app’s logs and health reports to audit decisions and refine how you blend automation with coaching.

Reps reframes the familiar readiness number into a decision engine that edits training plans before athletes open them. Its approach reduces the mental friction of daily training choices and surfaces concrete reasoning for recommended changes. The constraints and trade-offs are straightforward: Apple Watch dependency, limited third-party inbound integrations, and the balance between automation and agency. For athletes who value reproducible, morning-ready decisions and who already use Apple Watch, Reps offers a practical pathway from measurement to action.