MediPass

From a safe check today to medicine made for one person tomorrow.

The idea

One personal health record, kept by the patient, that an AI agent can use to pick and buy the right medicine at the right moment. It reads your medicines, allergies, conditions, gene results and live data from a ring or band. A doctor stays in the loop for prescriptions. Payments between agent, doctor and pharmacy happen automatically on Solana.

Why it is different

Other apps tell you what a medicine does. MediPass tells you what it does in you.

Personal, not general

A usual interaction checker compares one medicine with another. MediPass also uses gene results, so it can tell whether your body clears a drug fast, slowly or not at all. To our knowledge few consumer apps do this today.

At the moment of the decision

The warning appears when you buy or order, including over-the-counter medicines such as ibuprofen, which a prescription record does not contain.

One place, not four

Medicines, allergies, kidney values, gene results and your own notes sit together instead of in a doctor's system, a lab report, a drawer and your memory.

Only what is needed is shared

A pharmacist gets a short lived code with the warnings only, never the whole record.

One example, Max Mustermann, 58

All values are invented. The links between gene and medicine follow public guidelines such as CPIC.

Light	What MediPass finds	What it suggests
red	Clopidogrel after a stent probably works poorly for this patient, because the body clears it slowly (CYP2C19).	Ask the cardiologist whether prasugrel or ticagrelor would suit this patient better.
yellow	The muscle pain fits simvastatin together with the SLCO1B1 result.	Ask about a lower dose or another statin.
yellow	Ibuprofen adds bleeding risk with clopidogrel, can weaken ramipril and strains kidneys that already work less well.	Paracetamol is usually the better choice. Ask a pharmacist.
red	Amoxicillin, because of the penicillin allergy.	Do not take it. Ask for an alternative.

No single source shows this. The doctor does not see the gene, the pharmacy does not see the ibuprofen, and the patient thinks the muscle pain comes with age. Joining the sources reveals two possible treatment problems and an explanation for the complaint.

Where the data comes from, and its limits

Sources

The patient record (medicines from e-prescriptions, letters, diagnoses, lab values such as kidney function), a one off gene test that covers around ten to fifteen genes, and the person's own notes on over-the-counter medicines, effects and side effects.

Limits we respect, to our current understanding

Health and gene data have special protection under the GDPR and need explicit consent, so the data should stay on the person's own device. In Germany, genetic tests may only be ordered by doctors, so MediPass takes in the result and never orders a test. Third party apps cannot simply read the national patient record, as access runs through the insurer's app. For the start, people enter their medicines by hand or scan their medication plan. These legal points were not freshly verified, so check them before relying on them.

Where we are today

Built in this hackathon	How it works
Traffic light check	Rules for interactions, gene results (CYP2C19, CYP2D6, DPYD, SLCO1B1, CYP2C9), allergies and conditions. The verdict is never decided by an AI model.
Care agent	Understands a free sentence in any language, such as "I have a very strong headache, get me tablets". It checks red flags first, ranks the options against your record and your ring, and orders the safest one. The AI model reads the sentence and may only add caution. The rules decide.
Medication review	Checks everything you already take against your genes, allergies and kidney values, worst first, with a concrete next step and a hint when a complaint may be explained by a medicine.
Doctor in the loop	Prescription medicines wait for a quick yes or no from a doctor who sees the record and the check. No appointment.
Agent payments	The agent pays the pharmacy and the doctor on Solana devnet with test USDC. Where no receiving address is set, the payment is clearly labelled as simulated.
Pharmacy code	A signed 15 minute code that shows warnings only, never the profile.

Honest limits. The rule set is a demo and not a validated database. Ring readings are simulated. Orders and payments use test money. No real prescriptions are issued.

Two kinds of medicine, two kinds of personal

Small molecules

Aspirin, ibuprofen, clopidogrel

You cannot redesign a small molecule for one person, because a changed molecule no longer fits the enzyme and stops working. What you can personalise is the choice and the dose. Gene results such as CYP2C19 or DPYD predict who clears a drug fast or slowly. This is where MediPass works today.

Antibodies and mRNA therapies

Made for one patient

A tumour is different in every person. Its mutations can be read, and a therapy can be built against them. Moderna and Merck report that their individualised neoantigen therapy intismeran autogene met its main goal in a Phase 3 melanoma trial in August 2026. This is the long term bridge, and it is not cheap or fast yet.

"Personalised medicine" means different things here. The method depends on the drug type, so we keep the two paths apart.

Four year plan

2026 to 2027

Check and order

Pilot with a few pharmacies and doctors. Over-the-counter orders and doctor approved prescriptions. Rules reviewed by pharmacists. Regulatory path clarified for medical software.

2027 to 2028

Patient owned data

Connect rings and bands, labs and gene reports with consent. Pay per use between agents and services instead of selling the data.

2028 to 2029

Closed loop

Track whether a medicine worked and whether side effects appeared. Use it to improve advice and doses for the person.

2029 to 2030

Bridge to made to order

Partner with makers of personalised therapies. MediPass becomes the consented data and ordering layer between patient, doctor and manufacturer.

This is a plan, not a forecast. It depends on regulation, partners and funding.

Market, with honest ranges

Segment	What the sources say
Pharmacogenomics	About 5.7 billion USD (2025) by one publisher up to about 19.6 billion USD by another, with growth of roughly 7.5 to 10.6 percent a year. Ranges differ because publishers count different things. Global Market Insights, Precedence Research
Smart rings	Medical rings alone are small, about 340 million USD in 2025 growing to about 461 million USD by 2030 in one estimate. Consumer ring forecasts grow much faster. 360iResearch, ABI Research
Individualised cancer therapy	First positive Phase 3 result for an individualised neoantigen therapy, reported on 19 August 2026. Nine more Phase 2 and 3 trials are running. AJMC

We did not build a bottom up revenue model today, so we show no single market number. Market reports are forecasts by publishers, and their definitions differ.

Who else is in this space

Company	What they do	What is different about MediPass
CUDIS	Smart ring with an AI coach, health data linked to Solana. Raised a 5 million USD seed round led by Draper Associates and says it sold over 20,000 rings (company claim). The Block	Collects wellness data and rewards wearing the ring. MediPass uses such data for the patient's own medicine decisions. A ring maker could be a partner.
Nebula Genomics	Offered genome sequencing where data could be shared with pharma, in launch era reports. Signed a first pharma sponsor in 2019. STAT	Data is the product. For MediPass the data stays with the patient and is used to keep them safe. We found no current information on how Nebula operates today.
Genomes.io	We could not verify this company in our research.	Not assessed.
Moderna and Merck	Build individualised cancer therapies from a patient's tumour.	They make the medicine. MediPass would be the consented data and ordering layer in front of such makers.

Policy proposal, a consented health data pass

The problem

Health data is scattered and incomplete. Companies that collect it often sell it on. At the same time people take medicines that interact with their own genes and other drugs, and AI agents are starting to buy things for us.

The proposal

A patient owned health record with a duty on every AI agent to check it before buying a medicine. Prescriptions need a human clinician in the loop. Every use leaves an audit trail, and the data is shared only for a stated purpose and for a limited time, as the pharmacy code already does.

Who is responsible

The clinician who approves a prescription. The operator of the agent for its safety checks and its logs. The pharmacy for dispensing. The patient for the accuracy of what they enter. Regulators certify the rules, as for any medical software.

Trade offs

Faster ordering must not mean lower quality of care, so red flags stop the order. Rules can be wrong or out of date, so every verdict names its source and says it does not replace a pharmacist. Rich health data can be misused, so we minimise what is shared. Cheap and fast access can push people to skip a conversation with a professional. Gene data can reveal family members, and access can be unequal.

MediPass does not replace advice from a doctor or pharmacist.