Almavivo: the on-device health platform

Almavivo

4 min read Original article ↗

The on-device health platform

The questions your doctor doesn’t have time to ask.

Almavivo runs the health-planning conversation that doesn’t fit in a 10-minute appointment. It weighs the human evidence honestly, builds you a personal plan, and keeps every answer on your device. No account, no server, no profile. We don’t even take your email address.

Your answers are built into a plan in your browser. No account, no upload, no third-party scripts. How we prove it →

Built on the evidence base of the NIH Office of Dietary Supplements, Cochrane reviews, and PubMed-indexed human trials.

For employers →
For clinics →
For practitioners →

Pledge
No account·No tracking·No third parties·No exceptions

On your device
The plan is built in your browser. Our server never sees it.

Not AI
Almavivo doesn’t use AI to generate your plan. How it works →

Example output
An Almavivo plan looks like this →

Example output

A plan, not a shopping list.

What an Almavivo report looks like at the end of the intake. Lifestyle anchors come first, supplements come last, chosen to support the rest, not replace it.

Every recommendation carries an evidence tier. Every exclusion carries a reason. Patterns the engine notices in your answers are named, so you can see why the plan looks the way it does.

Tier A

Strong evidence

Multiple high-quality human trials and meta-analyses converge on a clear effect.

Creatine. Iron in deficiency. B12 in deficiency. Plant sterols.

Tier B

Reasonable evidence

Supportive human trial data with some inconsistency or narrower indications.

Magnesium. Omega-3 EPA/DHA. Melatonin. Riboflavin for migraine.

Tier C/D

Mixed, traditional, or emerging

Smaller trials, narrower use cases, or long traditional use without modern RCTs. We label these honestly so you can decide.

L-theanine. Glycine. Ashwagandha. Reishi.

Most health quizzes

  • Store your intake answers from question one.
  • Build a profile attached to your name and email.
  • Run dozens of trackers, pixels, and ad-network beacons.
  • Generate recommendations on a server that sees everything you typed.
  • Sell, share, or merge your health data with other parties.

Almavivo

  • Almavivo never receives them. Every answer stays in your browser.
  • No account. No login. No profile.
  • Zero third-party scripts. Enforced by a strict browser policy.
  • The recommendation engine runs on your device. Nowhere else.
  • We have nothing to sell, because we don't have it.

The one moment data moves

Only if you choose to share it.

At the end of your plan, you can choose to print it, save it as a PDF, or send a structured summary to a health professional you select. Nothing happens automatically. Almavivo doesn’t send the email. Your own mail client does. You see exactly what is leaving, and where it’s going, before anything moves. If you don’t click send, nothing leaves your device. Ever.

Read how the consented share works →
01

Deterministic

Same inputs produce the same outputs, every time. No randomness, no temperature, no “creative” phrasing. The engine is a pure function over your answers.

02

Rules-based, not generated

Each supplement has a written rule: who it’s for, who it excludes, what evidence tier supports it, what dose, what interactions to flag. Nothing is invented at runtime.

03

Open source

The engine is published openly at github.com/almavivo/open-health-engine under Apache 2.0. Every question, every supplement rule, every exclusion, every safety guard, auditable line by line, with citations to NIH ODS, Cochrane, and PubMed-indexed human trials.

04

Runs on your device

The engine is JavaScript that ships to your browser and executes there. No server-side inference. No API call to a model provider. Your answers don’t leave the page.

05

No hallucinations, by construction

A rule can be wrong, and we’ll fix it when it is. But it can’t hallucinate a study, invent a dose, or fabricate a supplement that doesn’t exist. That failure mode isn’t available to a rules engine.

06

Safety guards are explicit

Pregnancy, anticoagulants, SSRIs, kidney disease, under-18, red-flag symptoms. Each is a written exclusion or clinician-review gate, not a probabilistic guess.

Your health is not data. It’s yours.

Your answers are not signals. They’re yours.

Your plan is not a profile. It’s yours.

For organisations

A wellbeing benefit your employees can actually trust.

Six tools built on one private intake: supplement plan, lab values in context, annual physical prep, medication review, shift-work plan, household profiles. Every answer stays in the employee’s browser. Employers never receive individual reports. Open source under Apache 2.0, auditable by your DPO.

See the employer brief →