Your assistant recorded examples; Relay itself is live

Open full screen Demo

What can I help with?

Relay is connected. When a question touches guidance from the organisations you chose, it comes with the answer.

A stand-in for the assistant you already use, such as ChatGPT, Claude or Copilot: Relay is the connector inside it, not the assistant. The guidance here is demonstration content, not published or endorsed by any body it describes. Nothing you type is stored by us, and with no model connected it goes nowhere else.

Trusted guidance, inside the assistant people already use

Relay is a connector. A person adds it to their AI assistant once, choosing the bodies whose guidance they want to carry: their regulator, their college, their employer. After that they ask what they were going to ask anyway, and when a question touches that guidance the reply carries it, with its source, where it stops, and who to speak to. Relay never turns guidance into advice, and where it has no source it says so rather than guessing.

What you are seeing

Each reply takes the shape a reply takes when an assistant has a connector: the assistant's own short answer, then a Used Relay row where it consulted Relay, then what the guidance adds, with its source, where it stops and who to speak to. Open the row to see exactly what Relay returned. Turn Relay off from the + menu to see the assistant on its own.

No live assistant model is connected to this copy of the demonstration. For the suggested questions, the assistant's words were written in advance by Claude from exactly what Relay returns, and are labelled Recorded example; the Used Relay row is Relay answering live on every press. A question you type yourself gets Relay's live result.

What is real, and what is not

The assistant's answer

For the suggested questions, written in advance by Claude from what Relay returns, and labelled as a recorded example. For your own questions, not shown until a model is connected.

Relay's guidance card

Real: the same retrieval, citations, limits and escalation the connector runs.

The guidance in the connectors

Demonstration content written by Humanity Incorporated and marked on every citation. The doctors' material paraphrases GMC guidance; it is not the GMC's wording and the GMC has not reviewed or endorsed it. Each citation links to the GMC's own page, and the GMC documents it draws on are listed in the GMC source catalogue. Relay @GMC sets out exactly what it paraphrases and how the GMC could take it over.

Your message

Not stored by us or by Relay. It goes to OpenAI to produce the assistant's answer. Leave out anything that identifies a patient, a pupil or a colleague.

Try it for real, in your own assistant

The real thing needs no demonstration page: Relay is a connector you add to the assistant you already use. Open a join page below, follow it once, and it gives you a connector address. Add that as a custom connector in your assistant's settings (Claude and ChatGPT both support custom connectors, on most paid plans), then ask your question there. Your own assistant decides when to use it, exactly as Claude does here.