What is an A2A Agent Card, and does a website need one?
A small JSON file at a well-known address that describes an AI agent: its name, what it can do and how to connect to it. It is meant for sites that run an agent other agents are supposed to call, which most websites do not.
What it actually is
A2A, the Agent2Agent protocol, lets one AI agent hand work to another. Before it does, it needs to know who it is talking to. The Agent Card is that introduction: a business card, published at /.well-known/agent-card.json, listing the agent's identity, its skills, the address to reach it and how to authenticate. An older name for the file, agent.json, is still in use and we read it too.
Why it matters
If you operate an agent that others should be able to find and call, the card is how they discover it without being told in advance. Without one, a calling agent has nothing to read and no way to learn what yours offers. For a publisher with no agent to offer, there is nothing to publish and nothing is lost.
What good looks like
Only if you run an agent: a card that names it, lists the skills it genuinely has, and points at an endpoint that answers. A card with a name and nothing else describes nothing, so we do not count it. For everyone else, no card is the correct state.
Being honest about it
This is an Advanced check. It never fails a page and stays out of your headline score, because a missing card is expected for nearly every site. We keep it because agents calling agents is a real and growing pattern, and it is a signal worth seeing when it is there.
Where most sites go wrong
The common failure is a card left over from a demo: a file that names an agent that no longer exists, or an endpoint that has moved. The other is a site that answers every well-known address with its ordinary page, so a careless probe would think a card exists. We treat an HTML page at that address as no card.
On the web
In October 2026, 1% of the sites we scanned passed the “A2A Agent Card” check.
| Result | Share of sites |
|---|---|
| Pass | 1% |
| Worth a look | 89% |
| To fix | 0% |
| Couldn't judge | 9% |
Next.js sites pass it most often, at 4%. WordPress sites pass it least often, at 0%.
Not enough history yet.
Based on the sites Bot Appétit has scanned: ones people chose to scan, plus lists we scan for research. Not a random sample of the web.
Not sure where yours stands?
A free scan checks this and everything else in about 20 seconds — no signup. Or if it's already flagged and you'd rather it were simply handled, we do that too.
Flagged for this one? We handle it. No pitch, just a plain-English chat.
Book a call