Where the numbers come from
Dosavy states doses, half-lives and schedules. A site that does that owes you an account of who writes them down and against what.
What Dosavy is
Dosavy is tracking and reference software for peptide, GLP-1 and TRT/HRT protocols. It records what you tell it, does the arithmetic, and cites published sources for what the library states.
I started by asking the AI assistants. Most of the time they would not answer at all; a dosing question got a paragraph about consulting a professional and nothing else. When one did answer, it gave me a number with real confidence and no way to tell where it came from: a trial, a vendor page, a forum post, or nothing. I could not use either kind of answer, and I could not tell which was worse.
So I built the library the other way round. No number goes in without the paper it came from, and the paper is linked next to it, not summarised. Where the studies exist but none measured a dose in people, the entry says so instead of borrowing one. Where there is essentially no literature, it says that too. As of today, 18 of the 50 compounds carry one of those two labels. 15 compounds are prescription-only or controlled and are in the app without a public page, deliberately, whatever their evidence status.
Dosavy is that library with three things reading from it: a reconstitution calculator, a dose log, and a question layer. They share one implementation, so the calculator and the app cannot disagree about a draw volume, and none of the three can state a number the library does not hold. When it has nothing, it says there is no published human dose for this, and stops. That sentence is the reason the whole thing exists. Dosavy is run from Fribourg, Switzerland; it sells no compound and takes no money from anyone who does.
How the library and calculators are sourced
The compound library covers 50 compounds, curated by hand from published literature, regulator labels and clinical trial reports, cited by PMID. Every compound carries an evidence label naming which of those it rests on: an FDA or EMA label, a clinical trial, or no human dosing evidence at all, which is the honest state of 18 of the 50, and the library says so rather than filling the gap with a number.
The AI advisor is built on one rule: it never states a figure that did not come back from a row in that library, citation attached. Where the library is silent, the advisor says so.
The public calculators follow the same standard. Each page writes its formula out in plain text so you can check it by hand (the app runs the same functions, not a second implementation that drifts), and numeric claims on those pages carry their source inline: FDA labeling, USP <797>, WHO standards, or the cited trial.
There is no clinician on staff and no medical review board, and no page on this site claims otherwise. What we can promise is narrower and checkable: every stated number traces to a named published source, and anything that cannot is either labelled as community practice or not stated.
How we grade evidence
We grade the human dosing evidence for each compound, not the number of papers written about it. Forty rodent studies and no human dose-finding trial is No human dosing data; one FDA label is Cited human doses. A count of references says how much has been written. The label says whether anyone has measured a dose in a person.
- Cited human doses
- A human dose has been published, by a regulator or in a trial, and the range on the page is that one, cited.
- No human dosing data
- Studies exist, often many, but none measured a dose in people. The page shows the vial math and no dose.
- Not studied
- Essentially no published literature. Nothing to grade, and the page says so.
- Animal data only
- Animal or in-vitro work only. It can explain a mechanism; it cannot set a dose.
- Limited human data
- Human data without randomisation: small studies, case series, open-label work.
- Randomised trials
- At least one randomised controlled trial in humans.
The two labels are not one ladder. A regulator label or any published human trial can yield a cited human range; the evidence level separately records whether that human data was randomised, so a cited range can sit next to limited human data. Animal work explains a mechanism and never yields a range. Where sellers and clinics list a figure nobody has trialled, it is labelled as observed practice and kept apart from the evidence.
What Dosavy is not
Dosavy is not telehealth and not a medical device. It does not prescribe, dispense, sell or diagnose, and it is not a substitute for a qualified clinician. It is informational tracking and reference software, and the Terms of Service state that boundary in full.
Who operates Dosavy
Dosavy is operated from Route du Musée 2, Fribourg, Switzerland. The same operator is named as responsible in the Privacy Policy and Terms.
Corrections are part of the sourcing promise: if a number on any public page is wrong, unclear or missing its source, email support@dosavy.com and it gets checked against the literature, not defended.
Where your data goes
- Operator
- Dosavy, Route du Musée 2, Fribourg, Switzerland.
- Commercial ties
- None. We sell no compound, link to no seller, carry no affiliate link, take no sponsor. If that ever changes, this line changes first.
- Where data lives
- Supabase, European Union (Frankfurt). Email reminders go out through Resend, European Union (Ireland), discreet by default: no compound name in a subject unless you turn that on.
- Analytics and errors
- In your browser, nothing runs until you consent: no analytics, no error reporting, no session recording. PostHog, European Union (Frankfurt). Our servers separately count a fixed list of product events against your account id, and stop if you decline. No event carries a compound name or a dose. Error diagnostics from our own servers are not consent-gated; they carry no cookies, request bodies or query strings (Privacy Policy, Section 2).
- The question layer
- The part of the app that answers from the library is opt-in behind its own consent, and off until then. What leaves: your question, the curated library rows it needs and, when you ask about your own regimen, the protocol summary described in Section 7 of the Privacy Policy (doses, schedule, adherence counts). What never leaves: your name, your email, your protocol names, your notes, your injection sites, your dose-by-dose log (it receives counts, not the log). It is received by a third-party model provider, named in Section 9 of the Privacy Policy.
- What is kept
- Every exchange with the question layer, for 5 years, so any answer can be audited after the fact. Deleting your account deletes everything else in one transaction.
- What it cannot do
- State a number the library does not hold. 18 of 50 compounds return no dose. A prescription-only compound is answered with its regulatory status first, and any published range is reported as what the literature says for prescribed use, never as a dose for you.