3D illustration of a padlock, a shield and a document with highlighted lines, representing secured agent access to clinic bookings

Research pilot · Not publicly live

Help shape how patients may book through AI assistants.

Vascue is researching a patient-authorized booking layer for assistants such as ChatGPT and Claude. We are starting with clinic interest and synthetic testing, not live patient bookings.

Cliniko first
Synthetic testing
Clinic approval

How it works

From “book me a physio” to a confirmed appointment

A patient asks their AI assistant

“Find me a physiotherapist near Central and book Thursday afternoon.” Their assistant goes looking, and finds your clinic.

Your clinic shows up with real openings

The assistant sees your actual services and open times, and offers them to the patient on the spot. No phone tag, no forms.

The booking lands in your calendar

The patient confirms a one-time code, and the appointment is written into the system you already run: a verified booking, no double entry.

Pilot design

A new front door, with the locks already on

Patient authentication first

The pilot design requires the patient to authenticate outside the agent conversation before protected booking actions are available.

Synthetic testing first

The first evaluation uses test patients and a test calendar. Live clinic access requires a separate review and approval.

The clinic stays in control

A future pilot begins with narrow services, explicit permissions, and clinic review before expanding any workflow.

Patient data, handled with care

Vascue is ISO 27001 certified. Privacy, security, and provider reviews are release gates before any real-patient pilot.

Built by the team already running healthcare AI at scale

150,000+

messages handled

99.5%

replies rated correct by staff

~60s

from referral photo to quote

8,000+

patients served

Live in a private hospital radiology department in Hong Kong. Figures as of August 2026. Read the case study

FAQ

What clinic owners ask first

Is this live today?

No. This is a controlled research pilot, not a publicly available booking service. We are collecting clinic interest and testing with synthetic data before considering any real-patient launch.

Which AI assistants can book us?

The research is designed around assistants such as ChatGPT and Claude. Any provider would require technical testing, privacy and security review, and explicit approval before accessing protected booking tools.

How do you stop fake or spam bookings?

The proposed design requires patient authentication, explicit confirmation, clinic approval, rate limits, and server-side authorization before a booking can be written. These controls are still being tested.

Do we have to change how we work?

No production connection is made from this form. If the project resumes, the first pilot will use a narrow Cliniko workflow and clinic review before anything reaches a live calendar.

Research notes

The agent-ready healthcare series

How the research pilot is built: what an MCP booking server exposes, why agents never touch clinic databases, and the guardrails between a patient's assistant and a confirmed appointment.