Everything a private prescription needs, and nothing it does not.

Every medicine you add is normalised against a vocabulary of more than 2,100 drugs, assembled from the openFDA product directory and a hand-curated, citation-backed core. The check then runs on the whole draft at once and returns three things.

• Drug-to-drug interactions, each graded minor, moderate, major, or contraindicated, and each carrying its mechanism, the advised action, and a source you can read.
• Allergy conflicts, matched by drug class against the patient's recorded allergies, not only by exact name.
• An unmatched list. The result is coverage-aware: it says which names it could not resolve, so a quiet result is never mistaken for a clean bill.

The same interaction data powers the free, no-account checker on Docuity Home, if you want to see the engine before you sign in.

Pick the patient, then add each medicine with its dose and directions. Autocomplete draws on the full vocabulary as you type, and the safety net re-runs whenever the list changes, so what you see always reflects the current draft. Nothing is checked in the browser: every lookup is proxied server-side.

Issuing moves a draft to a signed, immutable record. Before that can happen, every one of these must hold:

• the prescription is still a draft;
• the person issuing has a prescriber role;
• a medical-council registration number is on file for them;
• there is at least one medicine on the script;
• none of the medicines is a controlled substance.

The controlled-substance block is independent of the interaction dataset. It runs a conservative offline deny-list against the raw drug name at issue time and fails closed, so a scheduled substance cannot slip through during a service outage. Controlled scripts are simply out of scope for v1.

An issued prescription can be printed as an A5 private-prescription sheet, exported as a PDF, or transmitted to a pharmacy your practice has invited. Pharmacies join by invitation only and work issued scripts from a shared inbox, moving each through received and dispensed.

The lifecycle is explicit: draft → issued → sent → received → dispensed. A prescriber can cancel at any point before dispensing.

Every issued script carries a QR code and a short verification link. The public verify page returns four facts and no more: the status, the issuing practice, the date, and the item count. Patient identity and medication detail never leave the server, and the page is rate-limited.

Prescription lifecycle events are written to an append-only, hash-chained audit log whose database permissions do not allow updates or deletes. Chaining makes tampering detectable by verification. We describe it as externally verifiable, never as tamper-proof.