Free tier available / Launching 2027

Intelligent health guidance, built into your website.

Vertex Health helps patients understand their symptoms, ask the right questions, and find the appropriate next step, from self-care and OTC guidance to connecting with a healthcare professional. Free to embed, and never billed to your patients.

The assistant is already running on this page. Open it in the corner and try to break it.

Sample sessionAbbreviated
Patient
I've had a heavy pain in my chest since this morning and now it's going down my left arm.
Engine

Rule cardiac.chest_pain matched. Outcome forced to EMERGENCY.

Assistant
Stop what you are doing and call your local emergency number now. Do not drive yourself, and do not wait to see if it passes.

The model did not make that call. A deterministic rule did, and the model cannot argue with it.

Product
White-label AI health assistant
Built for
Healthcare businesses
Price
Free, or $199 a month
Status
In development
Launching
2027

01How it works

Three steps, one question at a time.

  1. 01

    The patient says what is wrong

    In their own words, or by picking from a symptom list organised the way people actually think about their bodies: head, chest, skin, mood. Whatever language they write in becomes the language of the session, and stays pinned there so the questions never drift halfway through.

  2. 02

    The assistant asks what it still needs to know

    Each answer is stored as a structured clinical fact, so the next question is the one that actually narrows things down rather than the next item on a fixed form. Underneath it is an HL7 FHIR Questionnaire that rewrites itself as the picture fills in.

  3. 03

    The session ends in a decision

    Every conversation terminates in one of three outcomes below. There is no fourth, vaguer option, and the assistant will not leave a patient holding a maybe.

02Your branding

See it carrying your name before you sign up for anything.

This is the real widget, not a picture of one. Put your clinic's name and colour in and it repaints as you type. Ask it about a symptom and it will triage you for real, on the free tier.

#0f766e

Stays in your browser — nothing is uploaded. PNG, JPEG, WebP or GIF, up to 400 KB.

On the paid plan the escalation contacts and booking link are yours too, resolved server-side from your dashboard rather than from the script tag.

03Outcomes

Where a session can end.

  1. SELF_CARE

    Self-care and pharmacy guidance

    Named active ingredients like paracetamol, cetirizine and diclofenac gel, never brands, because the box looks different in every country and the ingredient does not. Contraindications like pregnancy are applied deterministically after the model has spoken, not by it.

  2. ESCALATE

    Hand off to your clinicians

    Your booking link, phone number and email, resolved server-side from your business ID. On the free tier there is no tenant to route to, so the patient gets generic advice to contact a local clinician instead.

  3. EMERGENCY

    Stop and call emergency services

    Triggered by red-flag rules for stroke signs, anaphylaxis, radiating chest pain and thunderclap headache. They run alongside the model and can override it. The patient is told to call their own country's number, not an American one.

04Safety and privacy

A missed emergency is the worst thing this product can do. Everything else is arranged around that.

Red-flag rules are code, not prompts

They are versioned, testable and they win. The rules can escalate a session on their own; nothing can talk them down. Thresholds are set deliberately low, because a false escalation costs a wasted appointment and the other mistake costs more than that.

The medicine list is an allow-list

The assistant can only suggest ingredients we have written down and reviewed. A hallucinated prescription drug has nowhere to go, because there is no path from the model's output to the patient that does not pass through that list.

Symptom data stays behind one boundary

No analytics SDK, no client-side error reporter and no third party ever sees it. Encrypted at rest, tenant-scoped on every query, and every read is written to an audit log. A cross-tenant leak would be a breach, so tenancy is enforced at the query layer rather than remembered by hand.

The free tier cannot run away with itself

An unauthenticated endpoint that reaches a paid model is a standing risk, so it sits behind per-IP and per-origin rate limits, a signed short-lived session token, Turnstile, and a daily token cap. When the cap trips, triage degrades to safe escalation rather than failing open.

Where we actually are

Plainly: the red-flag rules and the over-the-counter list are a first pass. They have not yet been validated against a licensed triage protocol or signed off by a pharmacist, and the BAAs are not signed. All of that lands before a single real patient uses this. That is why the date says 2027.

05Languages

Patients get answers in the language they asked in.

The widget ships curated interface strings, and the model replies in whatever language the patient writes. One language is pinned per session and re-pinned only when the patient switches on their own, so a conversation cannot start in Spanish and finish in English. The dashboard and this site are localised too.

Interface
العربية·Беларуская·Български·বাংলা·Català·Čeština·Dansk·Deutsch·Ελληνικά·English·Español·فارسی·Suomi·Français·עברית·हिन्दी·Hrvatski·Magyar·Bahasa Indonesia·Íslenska·Italiano·日本語·Қазақша·한국어·Монгол·Bahasa Melayu·Norsk bokmål·Nederlands·Polski·Português·Română·Русский·Slovenčina·Српски·Svenska·தமிழ்·తెలుగు·ไทย·Türkçe·Українська·Tiếng Việt·中文
Conversation
Any language the patient writes in

06Pricing

Two plans. One of them is free.

Free

$0No account, no card

  • Vertex Health branding, with a small badge in the corner
  • Escalation gives generic advice to see a local clinician
  • Every safety rule the paid plan has, with nothing held back

White label

$199Per month

  • 500 completed assessments a month, then 10 cents for each one after that
  • Your logo, your colour, your name, and no Vertex Health badge
  • Escalation routes to your own clinicians, booking link and phone

A completed assessment is a session that reaches one of the three outcomes above. Conversations a patient abandons partway are never counted and never billed.

07Install

One line of HTML, before the closing body tag.

Free, no account

Copy this and you are done. It loads the assistant into an isolated iframe, launcher included, so nothing of ours touches your styles or your DOM. This is the exact embed running on this page.

<script src="https://www.getvertexhealth.com/widget.js"></script>

White label, on the paid plan

Adding your business ID resolves your colours, logo and escalation contacts on the server. Branding is deliberately not a script-tag option: nobody should be able to wear another clinic's identity by editing HTML.

<script src="https://www.getvertexhealth.com/widget.js" data-business="12345"></script>
Cost
Free$0
White label$199 a month
Completed assessments
FreeRate limited, not metered
White label500 a month, then $0.10 each
Account
FreeNone
White labelRequired
Branding
FreeVertex Health, with badge
White labelYour logo and colour
Escalation goes to
FreeA local clinician, generically
White labelYour clinicians

Script-tag options

data-hide-on-pages
Comma-separated path patterns where the widget should not appear, e.g. /checkout,/admin/*.
data-default-language
Initial language. The assistant still replies in whatever language the patient writes in.
data-dark-mode
auto, light or dark. Defaults to auto, which follows the patient's system setting.
data-position
bottom-right or bottom-left.
data-z-index
Stacking order of the launcher, for sites with fixed headers.
data-auto-open
Delay in milliseconds before the assistant opens on its own. Omit to keep it closed.

Colours, logo, contact and booking details are not in this list on purpose. They live in the dashboard and are resolved server-side from your business ID.