The agent layer, documented.
How patients' AI assistants find, talk to and book the practices on your platform.
Patients have started asking their AI assistants to handle real tasks, and booking an appointment is one of them. The agent layer is what makes the practices on your platform reachable when that happens.
It works like this: every practice gets its own AI reception endpoint under your brand. A patient's assistant (ChatGPT, Claude, Gemini or anything that speaks MCP or A2A) discovers it, talks to it, and books a real appointment that lands in your backend as a real calendar write. The patient confirms who they are with an SMS code inside the conversation. No patient account, no app, no portal.
What you integrate once, works everywhere
One integration into your backend powers every door:
- MCP for assistants that use tool connectors (claude.ai, ChatGPT and the growing list of MCP clients)
- A2A for agent-to-agent protocols (agent card discovery + JSON-RPC messaging)
- Phone and WhatsApp reception run on the same integration, included
The shape of the system
| Piece | What it is |
|---|---|
| Agent card | A public JSON file per practice that tells any agent what it can do here |
| MCP endpoint | One conversational tool: reception_message. The assistant talks, the reception answers |
| A2A endpoint | The same conversation over the A2A JSON-RPC protocol |
| Your backend | Where availability is read and bookings are written, in real time |
Everything is scoped by a per-practice public token in the URL. The token identifies the practice; it never authenticates the patient. Identity comes from the SMS verification step inside the conversation.
Who this is for
Practice-management platforms (PMS/PVS) in the UK and DACH that want their practices bookable by patients' AI assistants without building agent infrastructure themselves. It runs under your brand, on your invoice. We never sell to practices directly.
Try it before you read further
There is a live demo practice. Add this connector to Claude, ChatGPT or any MCP client and ask it to book you a physio appointment:
https://founderexecute.com/api/a2a/a2a_4adce6dd8c149abd94d885bbe7a32323cfc9745edfa53b8e/mcp
Its agent card is at https://www.founderexecute.com/api/a2a/a2a_4adce6dd8c149abd94d885bbe7a32323cfc9745edfa53b8e/agent-card.json. The practice is ours and the calendar is a demo calendar, so book freely. Everything documented below is what that connector is doing.
Start with the agent card, then the MCP endpoint, then what we need from your backend.
Put your platform on the agent web.
One integration call: your backend, the channels, the commercials — and a live booking you can watch.