Questions
Answered directly, including where the honest answer is that it depends or that nobody can tell you from the outside. Where a question genuinely needs a professional in your field, it says so rather than guessing.
Getting started
Other people's AI agents can find out what you do, ask you a question, and start a request such as a booking, without scraping your web pages or guessing at your forms.
In practice that means three things: a public description of your agent at a standard address (an agent card), functions an agent can call on your website, and an endpoint that answers other agents. Each can be added separately, starting with the card.
No. The A2A specification describes how agents discover each other and leaves registries to whoever runs them. Publishing your card at /.well-known/agent-card.json is the step that works everywhere.
Curated registries exist inside companies and platforms, and public ones are emerging. List your card in any registry the agents you care about actually search, but treat the well-known address as the foundation.
Publish an llms.txt file and an accurate agent card. Both are small, free, and useful even before anything else is built. The agent card builder here makes the card for you.
Then check the readiness list on this site to see which of the next steps, website tools or an A2A endpoint, fits how your customers already reach you.
Some do, and the number is growing, but nobody can give you a reliable figure for your market. The groundwork, structured services and clear rules, is worth doing for its own sake.
Be wary of anyone quoting precise adoption numbers. What is certain is that the protocols are settling and the cost of publishing a card is close to zero.
Your website
It can request one, if your site offers a booking function to agents. Design it so the agent proposes a time and your business confirms it.
Under the WebMCP draft, a page registers tools such as list_services, check_availability, and request_booking. Browser support is early, so the page must keep working normally where the API is missing.
No. A chatbot is an assistant you put on your site. Being agent-ready means your site works for the visitor's own assistant. You can do either or both.
The groundwork is shared: both need your services, prices, and availability as structured information rather than only prose.
Safety
Not without the person seeing the amount and approving it. Payment, cancellations, and anything that commits the business belong to a person or a written rule.
Design agent-facing functions as requests: an agent proposes, the person or the business confirms. Log every request so you can answer what happened later.
Yes, the same way they abuse a public form. Limit request size, rate-limit, validate every field, and require authentication for anything beyond public questions.
The A2A specification lets a card declare authentication schemes such as API keys, HTTP authentication, and OAuth. Use them for any skill that touches customer data.
About this site
Bridgepath AI Solutions publishes it. There are no affiliate links, paid placements, or vendor rankings; the only offer is Bridgepath's free readiness audit.
This site also practices what it describes: it publishes its own agent card, answers A2A requests, and lists its pages in llms.txt. Its own card uses A2A 0.3.0 today; an upgrade to 1.0 is planned.
Not answered here
If a question you have is not on this page, it is worth adding. The corrections page explains how to reach the editor, and questions that arrive more than once get added here.