Ask your clinical questions. Horizon queries official guidelines, scientific articles, your documents and the medical forum in parallel, then synthesizes a single sourced answer.
Tetracyclines are contraindicated beyond the 1st trimester and discouraged from early pregnancy[1]. For acne in pregnant women, benzoyl peroxide and certain topicals are preferred[2]. A dermatology reassessment is recommended[3].
Between one consultation and the next, you have a few minutes to decide. The answer is somewhere — in a guideline, a recent article, a drug-safety notice, a PDF you filed away last year.
But finding it means juggling ten tabs, cross-checking sources that contradict each other and rephrasing the same query three times. Too often, you decide from memory for lack of time.
Sources scattered across official guidelines, literature and your own documents
Generic AI answers, with no citation, impossible to verify
The nagging doubt: is this information still up to date?
Chat, documents, forum and multi-agent search — built for physicians, hosted in France.
For complex questions, a coordinator breaks down your query and dispatches agents in parallel across your documents, the forum, official guidelines and articles. One single, sourced, verifiable answer, reasoning streamed in real time.
French guidelines (HAS, ANSM, learned societies) and international articles (PMC, Lancet…) through a dedicated pipeline. Every answer cited.
Upload your PDFs: automatic extraction, hybrid indexing (vector + full-text), per-page citations. Scoped to your account or your team.
Drop a screenshot or a document into the conversation: automatic OCR, text used as context at once, no storing first. Perfect for a one-off question about a report or a lab result.
Ask the community, upvote the best answers, let the AI synthesize the threads in your specialty.
Discharge summary, referral letter, prescription or free-form synthesis: the AI drafts a structured document from the conversation, editable in the chat and exportable to .docx. Missing values stay flagged "to complete" — never invented.
AI model and data processed and stored in France at OVH, never with a foreign provider. Your interactions stay private: chat history is stored locally on your device. Per-team scoping and access logging.
No prompt to fine-tune. Ask the way you'd ask a colleague.
In plain language, the way you'd ask a colleague. No identifiable patient data required.
It breaks down your question and queries official guidelines, scientific articles, your documents and the forum in parallel — reasoning streamed in real time.
A single synthesis, every claim linked to its cited source. You always have the final say.
I built Horizon as the tool I wanted between two patients: a reliable, sourced answer without leaving my workflow or exposing any patient data. Not a gadget — an assistant that cites its sources and leaves the decision to you.
What every healthcare professional should know before getting started.
No. Don't include any identifiable patient data in your questions. Horizon is built for general clinical reasoning, not for handling named records. The AI model and data are hosted in France, at OVH, with per-account and per-team scoping.
Every search answer is grounded in real, cited sources: HAS/ANSM guidelines, learned societies, PubMed/PMC articles, or your own documents. You can trace each one back to its source in one click before you decide.
Never. Horizon synthesizes and sources information to save you time. The decision, the interpretation and the responsibility remain entirely yours.
A few minutes. Create your account, ask your first question — no installation, no credit card required to start.
French guidelines (HAS, ANSM, learned societies), international literature (PubMed, PMC, The Lancet…), your uploaded documents and the peer forum — in parallel, then synthesized.
Yes. Your uploaded PDFs are scoped to your account or your team, indexed for search and never shared outside your perimeter.
Create your account and get your first sourced answer today.
Start for free