Everything the app does, in the order you'll want it. Ten minutes end to end — or skip to the part you came for.
Updated 19 Aug 2026 · therapylog.app/guide
01Put it on your home screen
TherapyLog runs in your browser, and it installs to your home screen from there — real app icon, full screen, no address bar, dose reminders. There is no App Store or Play Store listing to hunt for, and nothing to download. That's deliberate: a store listing would mean cutting the compound information this app exists to provide.
iPhone & iPad
Open therapylog.app/appin Safari. Chrome on iOS can't install home-screen apps.
You never update it manually. The app checks for a new version each time you open it and refreshes itself in the background. If you want to force it, close it completely and reopen.
02Your license key
If you subscribed to Pro or BYOK, your purchase email contains a key that looks like TL-4RT9-K2MX-8QN3. Paste it into the app once and it unlocks on that device. Buying on the web usually activates itself the moment you land back in the app — the key matters later, when you move to a new phone.
AI → Settings → Activate a license. If the key isn't to hand, Use my purchase email instead looks you up by the address you paid with, and Email me my key sends it again.
Keep it somewhere you'll find it
Save the key in your password manager, or keep the purchase email. You'll need it when you replace or reset your phone — the key is how a new device knows what you bought. Losing it isn't fatal (we can look you up by your purchase email), but having it makes a new phone a thirty-second job.
One device at a time
Your log lives on the deviceActivating on a new phone is fine. The data doesn't follow the key, though — that's what backups are for.
Subscriptions
Checked automaticallyThe app re-verifies with Stripe periodically and shows your renewal date under Settings → Plan & Backup. Cancel any time from the link in your receipt.
03The five tabs
Everything lives behind the five buttons along the bottom.
Home — quick actions, refill countdowns, what's due today, and your streak.Health — labs, protocol calendar, progress photos, and trends over time.
Home
Today at a glanceDoses due, refills running low, quick log buttons.
Log
Everything you recordSix sections along the top: Weight, Meds, Body, Symptoms, BP, Labs.
Reference
130-compound encyclopediaMechanism, dosing ranges, half-life, side effects, interactions, monitoring. Plus Stack Builder, and Tools & Calc for dose and reconstitution maths.
Health
Labs and trends100 bloodwork markers charted, with ranges personalised to your age and sex. Also the protocol calendar and progress photos.
AI
The research assistantIts hub also holds Profile (where the clinical report is generated), History, and Settings.
04Logging a dose
Log → Meds. Pick the compound, the dose, and the site. The app remembers what you logged last time, so the second entry onward is two taps.
Site rotation is tracked for you — the app nudges you off a site you've used recently.
Reconstituting a peptide?Reference → Tools & Calc does the bacteriostatic-water maths and draws the syringe, so you can check the mark before you draw. Common vial presets (5mg/2ml, 10mg/1ml and the rest) are one tap.
Refill alerts come from your logged doses and vial size. Log consistently for a week and the countdowns on Home become accurate.
Missed a day? Back-date it. Tap the date field at the top of the entry.
05Getting your labs in
This is the part people underuse. Log → Labs. You do not have to type a hundred numbers.
Photo, screenshot, or PDF — all three work. Upload pulls from your files, not just the camera, so a PDF emailed by the lab goes straight in.
The scanner
Add every page. A twelve-page LabCorp PDF can go in as one file; a phone photo set can go in as twelve images. Queue them all, then scan once.
It reads the lab's own reference range off the page and uses that instead of ours whenever it's printed — your lab's range for your assay beats a generic one.
It won't guess. A marker it can't confidently identify is listed as unrecognised rather than mapped to something that looks similar. That's on purpose: a wrong marker is worse than a missing one.
Check before you save. The values land in the form for you to eyeball. Fix anything that looks off, then save.
Typing it in
The full form covers 100 markers. Search it, fill only what you have, and leave the rest blank — blanks aren't treated as zeroes.
Anything the scanner missed, type in. Same form, same trends, no difference downstream.
Units are switchable per marker — ng/dL or nmol/L for testosterone, pg/mL or pmol/L for estradiol. Pick what your lab printed; the app converts for the charts.
Assay matters and the app asks. Sensitive (LC/MS-MS) estradiol and standard immunoassay estradiol are not the same measurement, and mixing them makes a trend line lie. Where it matters, there's a picker — set it to what your lab ran.
A marker we don't have? Add it as a custom row. It charts like the rest.
06Reading your trends
Health → Labs. Green is inside range, red is outside. The range is personalised to the age and sex in your profile, so fill that in — it changes what counts as flagged.
Overview is the latest panel. Trends charts any marker over time. Log Labs is the entry form again, and History is every panel you've entered.
Optimal bands are shown separately from your lab's reference range, and they are not diagnostic — they're where a lot of practitioners aim, which is a different question from whether a value is abnormal.
Two panels is a trend, one is a data point. The charts get useful at the second draw.
Clinical report — AI → Profile → Generate Clinical Report — turns your log into something you can hand to a doctor: protocol, dose history, lab trends, body composition, adherence. Generate it the night before an appointment.
07Asking the AI well
The assistant already knows your logged protocol and your labs. You don't need to paste your numbers in — ask about them directly.
Answers lead with the point and stop. Copy and Export put an answer somewhere you can keep it.
Two things are worth knowing about how it's built, because they change how you should ask:
It searches before it states. It can run up to three literature searches per question, against PubMed, ClinicalTrials.gov, DailyMed, Cochrane, the major journals and a set of clinical references. Ask it to cite and it will.
It answers short by default and offers to go deeper, because most questions asked from a phone want the answer, not an essay. If you want the essay, say so in the first message.
Ask for the depth up front
Thin
tell me about tesamorelin
You'll get three tight paragraphs and an offer to expand. Fine if that's all you wanted; a wasted question if it isn't.
Deep
Search the literature and give me the full picture on tesamorelin: mechanism, documented dosing ranges, half-life, the realistic side-effect and risk profile, key interactions, and the bloodwork I should be monitoring. Label every claim as established clinical use, off-label or community practice, or animal-only. Cite what you find.
Names the sections you want, asks for the evidence tiering, and triggers the searches. This is the single highest-value habit in the whole app.
Use your own data
Grounded in your log
Look at my last three lab panels and tell me which markers are trending in a direction I should be watching. For each one, say whether the change is likely protocol-related, and what I should raise with my prescriber.
No numbers pasted. It reads your panels itself, and it tells you which markers were not tested rather than inventing them.
Plans, not just facts
Training
Build me a 4-day upper/lower program around a healing left rotator cuff. Commercial gym, 60 minutes a session, 8-week progression. Include the movements to avoid and why, and how to know I'm pushing the shoulder too soon.
Constraints — days, equipment, time, injury, length — are what turn a generic plan into a usable one.
Nutrition
I'm 5'11", 205lb, on a GLP-1 at 1mg weekly, lifting 4×/week. Give me a protein and calorie target for preserving lean mass while losing fat, and three days of example meals that hit it. Flag anything that gets harder on a GLP-1.
Give it your stats and your protocol and it works from them rather than from an average person.
Follow up narrowly
It remembers the conversation, so follow-ups can be short. Good ones: "Which of those claims are animal-only?" · "What's the monitoring schedule for that?" · "Give me the study details on the second point." Use Clear history when you change topic — a stale thread pulls answers back toward the old subject.
What it won't do
It won't invent a dose or a study, and it says so when the evidence is thin. It won't help you synthesise or home-brew raw materials. And it won't prescribe — for anything injected, any dose change, any protocol decision, it lays out what's known and tells you to set it with a provider and bloodwork. That's the deal, and it's why you can trust the rest of what it says.
08Making your questions count
On Pro you get 50 questions a month, resetting on the 1st. One send is one question, however long it is — so the shape of your asking matters more than the length.
One send
= one questionA 300-word question with five parts costs exactly the same as "hi".
Follow-ups
Also cost oneWhich is why asking for the depth in the first message is the whole trick.
BYOK
UnlimitedBring your own Anthropic key and there's no counter at all — you pay Anthropic directly for what you use.
The habits that stretch it
Bundle. Number your sub-questions in one message instead of sending four. Same cost, and the answer is more coherent because it sees the whole picture.
Say how deep, and in what shape. "Full mechanism, dosing by evidence tier, monitoring schedule, cite sources" gets you in one question what three rounds of "tell me more" would cost.
One or two compounds per question. An answer has a length ceiling and only three searches to spend. Eight compounds at once gets you eight thin paragraphs; two gets you two good ones.
Let it read your log. Asking it to look at your panels is free and better than pasting numbers you might mistype.
Save the good ones.Copy or Export an answer worth keeping and you'll never pay for it twice.
Check the encyclopedia first. Dose ranges, half-lives, interactions and monitoring for all 130 compounds are already in Reference, free and offline. Ask the AI the questions the encyclopedia can't answer — the ones about you.
If you run out
You'll be told which limit you hit and when it resets. Adding your own Anthropic API key in AI Settings switches you to unlimited immediately, on any plan.
09Backing up
We never see your data. That's the point of the app, and it's also the one thing that asks something of you: your log lives on this device only. A lost, reset, or wiped phone loses it.
AI → Settings, under Plan & Backup. The app counts the days since your last backup and starts nagging when it's been too long.
Back up weekly. That's the recommendation and what the reminder is set to. Tap Back up now and pick your cloud — iCloud Drive and Files on iPhone, Google Drive on Android, anywhere you like on a desktop.
Always back up before switching devices, and before a phone reset or an OS upgrade.
On a desktop, Set up automatic weekly backup links a file once and then rewrites it in place — no prompt after the first time.
Restoring is the same card, in reverse: pick the backup file and it loads. A restore on a second device gives you a copy of the log, not a live second screen.
10What it doesn't do
Worth knowing up front, so nothing surprises you later.
No account, no login. There's nothing to sign into because there's no server holding your data. Your license key unlocks features; it doesn't store a log.
No live sync between devices. Two phones means two separate logs. A backup moves a copy; it doesn't keep them in step. Cross-device sync would require us to hold your health data on a server, which is exactly the trade this app was built to avoid.
Not a medical device, and not medical advice. Flags, optimal bands, and AI answers are informational. Dose and protocol decisions belong with a qualified provider and real bloodwork.
Lab files aren't stored. A scanned report goes to Claude for reading and isn't kept by us. The values it extracted stay on your device.