The Lab Agent reads each new blood test, compares it with your last one and your DNA, checks the research, and keeps a plan you can act on.
You tap to run each new lab; a person starts every run. Each plan item is marked Cited, Hypothesis or Ask your clinician, with a dated retest. Next time you upload a lab, it checks whether the plan worked and changes it. It never diagnoses and never suggests doses or treatment changes; anything clinical becomes a question for your clinician.
The demo uses a fictional person's synthetic data. No sign-up, no password; your copy is deleted after 7 days.
The steps are fixed in code and always run in the same order, after you tap to run a new lab. A language model judges the papers and words the plan; fixed rules decide what the agent does.
Pulls the text out of the PDF your lab sent and picks out the markers it knows, such as vitamin D, LDL cholesterol or homocysteine.
Checks each marker against its reference range and against your previous result: in range or not, up or down.
Did the markers you worked on move, and did you keep the habits? Each habit is kept, swapped or retired based on that.
Looks for gene variants you carry that research relates to the changed markers. A link means "related", never "caused".
Searches PubMed and Europe PMC, and an AI judge checks each paper against each claim. A habit is marked "Cited" only when the papers support it; otherwise it stays a Hypothesis.
Small daily habits, a repeat test with a date, and questions for your clinician. Fixed rules route each item: Cited, Hypothesis or Ask your clinician.
Puts the habits, the retest date and a short sheet for your clinician into the app. Anything that leaves the app, such as a calendar or Telegram reminder, a clinician link or a lab order (US sandbox), waits for your tap and shows what it will share.




Screenshots from the live demo with synthetic data. Not a real patient.
The Lab Agent is a tab of the VitaCrypt app. The app keeps your health file and matches it to research; the agent checks each new lab and keeps a plan with you. Its plan shows up on Today, in Habits and in Trends.
| Without the agent | With the Lab Agent | |
|---|---|---|
| Memory | A snapshot of today's data | Keeps every lab, plan and habit it suggested |
| Plan | Recommendations to read | Habits in your tracker, repeat tests with dates, clinician questions |
| Follow-up | None until you ask again | Reminds you when a repeat test is due |
| Learning | Starts fresh each time | Checks the next lab against the plan and adjusts it |
The Lab Agent is not medical advice and is not a medical device. Every sentence the model writes passes a guard that blocks personal doses, medication changes and diagnoses; if it fails, a fixed template is used instead.
The live demo runs on a fictional person's data. Each visitor gets a private copy that is deleted after 7 days.
Only coded facts go to the language model (Anthropic's Claude), such as "vitamin D, low, steady" and a gene code. No name, email, exact values or file text. Zero data retention terms are requested from the provider, not confirmed.
The agent's step log records which step ran, when and how long it took. No values, no prompts.
Health data is stored encrypted under a per-user key held by our server, which unwraps it to run the agent. Client-held keys are roadmap. Deleting your data destroys that key.
Uploading your own lab results opens to beta testers in small groups, outside the EU and the UAE until a data protection review is done. The demo cannot upload files.
Follow Alex, a fictional person with a year of synthetic lab history: a new lab arrives, the agent explains what it will do, searches the research while you watch, plans habits and questions, and eight weeks later checks whether the plan worked.

The live demo needs no sign-up. Real lab uploads open to beta testers in small groups; the beta is free.