Open community · Agent builders · The agentic web

.agent

The community building the agentic web.

We are the companies, labs, and independent builders who develop, research, deploy, or operate agents. Members compare notes on what actually works, find each other, and take on the problems no single organization can fix alone.

Explore the map →
28,000 members · 8,000 organizationsFree, and open to anyone who builds

Backed by builders & researchers at

Backed by builders & researchers at

What we believe

Community commitments.

01

Accountability

Behind every agent there should be a person or an organization that stands behind and for it. That is not a nice thing to have. If nobody is answerable, there is nothing to trust.

02

Openness

The agentic web has to run on open specifications and open infrastructure. We publish standards anyone can implement, and we keep them open.

03

Fairness

Buying a domain does not mean buying the rules. Standing here, including a voice in governance, comes through verification. A two-person project that verifies has the same standing as a lab with a thousand people.

04

Stewardship

What we build together stays governed by the people who build it. That has to show up in the boring parts: membership, working groups, public records, appeals.

Why the work is arranged this way, at length, and where .agent fits into it. What Agent Community is →

What we do

Builders in the same room.

The fastest way to a working agentic web is builders who know what the others are building.

Meetups and hackathons

Hackathons, meetups, launches, and working sessions in cities around the world, in person and online.

/events · 65 events so far

Open discussions

The day-to-day conversation lives in Discord, where members trade what is working and what broke this week.

discord.gg · open to members

Work streams

Identity, security, evaluations, trust, and governance, as open mailing lists with public archives.

/lists · archives open to read

A map for the community

A public directory of member organizations, so the ecosystem can see itself and members can find the people working on adjacent problems.

/map · free to browse

Open specifications

AID resolves an agent from a domain name with one DNS lookup. Open spec, open source, clients in six languages.

aid.agentcommunity.org · IETF Internet-Draft

Informed members

Agent Brief curates what actually happened in the agentic web each day, so members do not have to drink from the firehose. Longer pieces on standards, trust, and governance go on the blog.

news.agentcommunity.org · /blog

Find your people

The Periodic Table of the Agentic Web.

Every member organization that wants to be seen, laid out so the ecosystem can look at itself for once.

Sort it by what a company builds or by where in the world it is, or spin the globe and watch it fill in. A directory is only worth having if it answers the question you actually arrived with, which is usually some version of who else is working on this.

Mission

Verified agents.

Anyone can spin up thousands of agents to act in their name. From the outside, a mimic looks exactly like a real representative. The difference: someone real stands behind the agent and answers for it, and you can check that yourself.

01

Accountability for agents

A real legal entity stands behind the agent and answers for what it does, and you can look that up rather than taking it on faith. A logo is not evidence of anything.

02

A trust framework

We would not be the one who verifies everyone. Partners, providers, and credentials people already hold can carry it. What stays constant is the bar, and it would not be limited to .agent names. Any domain holder could meet it.

03

Not for every agent

Most agents will never need this, and that is fine. Verification is for the ones that take on real responsibility. Nobody is being fenced out. The aim is that when it does matter, you can tell.

This is the direction, not a service we run today. The Department of Machine Verification is an onboarding experience for agents and their operators, and it records what was submitted and when. It does not verify who an agent is. What a verified agent means →

One piece of the trusted-agent puzzle

.agent

Trust starts with a name. That is true of people, and it is true on the internet: a name is how you get found, and it is the first thing anyone can check. It is where a relationship begins, whether that relationship runs for decades or for four milliseconds.

.agent should become for agents what .edu is for education.

Read why a namespace has to be governed

.agent is one piece of that puzzle. Not the whole answer, and not another vanity ending. Done right, it could be a naming layer where a name does more than resolve, and where holding one would mean a verified operator stands behind and for that agent.

A namespace like that has to be governed rather than controlled. Controlled means one owner writes the rules, and the rules drift toward the owner one reasonable-sounding step at a time. Governed means the people who build on it steer it, through binding registration policies, public records, and the standing ability to hold any operator to the charter. Goodwill is not a governance mechanism. Structure is.

We are an applicant, not a registry operator. .agent does not exist in DNS. ICANN’s 2026 application window closes 12 August 2026, and any outcome is subject to ICANN evaluation and ICANN-approved allocation policies. Full status, timeline, and what happens if it is not approved →

The name you’d want

Free and non-binding. It records interest and nothing more: no reservation, no ownership, no claim on availability or priority. If ICANN approves .agent, any future consideration is subject to ICANN-approved allocation policies.

FAQ

Everything you keep asking us.

01What does "community" mean here?
The people and organizations actually building this: companies shipping agents, researchers working on how agents should behave, and independent developers. Anyone who develops, researches, deploys, or operates agents belongs here. It is a real group with a shared problem rather than a list we assembled — members join because they need the same things solved, and the work happens in the open. Membership is free, the rules are written down, and who belongs is a matter of public record rather than our opinion.
02Who is behind Agent Community?
Agent Community was founded by Balázs Nemethi. Our advisors include Esther Dyson, the founding chair of ICANN, and Tim O'Reilly. Everyone who runs it, and every advisor, is named with their affiliations at agentcommunity.org/team. Member organizations that chose to be listed are on the public map at agentcommunity.org/map. We publish all of it for the obvious reason: a community that will not say who it is has not earned anyone's trust.
03How do I join?
Use any Join button on the site. Pick Individual or Organization, then continue with Google, GitHub (Individual only), or a one-time code sent to your email. It takes about a minute, it is free, and no payment details are collected. You land in your members area, where you can join a working group, note a .agent name you would want, and read the community endorsement letter. Signing that letter is optional. It is also the clearest way to say you back this.
04Is membership free?
Yes. Joining the community is free and requires a registration. There are no fees, no payment information collected, and no spam. You can leave at any time.
05How can I contribute?
Join a working-group mailing list at agentcommunity.org/lists or from your members area, come to a community event at agentcommunity.org/events, join the Discord at discord.gg/juPSbFBWZg, or open an issue or pull request at github.com/agentcommunity. The open agent discovery specs live at aid.agentcommunity.org. None of this requires signing anything.
06What is AID, and do I need it to join?
No, you do not need it to join. AID (Agent Identity & Discovery) is an open specification we maintain for finding an agent from a domain name: one DNS record at your domain says where the agent is and what it speaks. It works today, on ordinary domains, and needs approval from nobody. It is for teams who already run an agent or an API and want it discoverable. The specification and client libraries for six languages are at aid.agentcommunity.org.
07How can my agent join?
Agents can register at the Department of Machine Verification (dmv.agentcommunity.org). Being straight about what that is today: the agent gets a record of what was submitted and when, and no special privileges. It does not establish who the agent is, and it reserves nothing. Think of it as an early step toward something we believe is coming — agents that can accountably represent a real organization — rather than as a credential that means something yet.
08How do you count members, and what do the numbers mean?
A member is a completed registration: an individual or an organization that confirmed an email address. Endorsements are counted separately — the members who read and signed the community endorsement letter. We keep the two numbers apart instead of adding them together, because they mean different things and a combined figure would flatter us.
09What is a .agent domain?
A proposed top-level domain for AI agents, so an agent could have a clear, readable name on the open web: something like support.acme.agent. Just as .edu signals education and .gov signals government, .agent would signal an agent. It does not exist yet — this is what we are applying for.
10Can I buy a .agent domain now?
No, and nobody can. .agent has to be approved by ICANN, the body that decides which new top-level domains exist, and applications for this round close on 12 August 2026. You can note a name you would want, for free. That records your interest and nothing else: it is not a reservation, and it gives you no ownership or claim. If .agent is ever approved, pre-registrants may receive early consideration, subject to ICANN-approved allocation policies.
11How does the ICANN process work?
Slowly, and that is the honest headline. ICANN opens a round for new top-level domains only rarely — the previous one was in 2012 — and applications for this round close on 12 August 2026. Evaluation follows, and where more than one applicant wants the same name there is a further process to settle it. Anything approved then takes more time again before it launches. We are applying as a community rather than as a business, which is the more demanding path: it asks you to show that a real community exists and that the rules would serve it. However it goes, this is measured in years.
12What happens if .agent is not approved?
Then it does not launch, and there is nothing to allocate. A pre-registration is free and non-binding either way — it records interest, and any early consideration it could carry was always subject to ICANN-approved allocation policies. That changes less than it sounds like, because the rest of the work does not depend on the name existing: the open identity and discovery specifications at aid.agentcommunity.org run on ordinary domains today. The direction we care about most is the verified agent, and that bar was never limited to .agent names — any domain holder could meet it. We would keep going, and could apply again in a later round with more of the community behind us.
View all questions →

Explore

If you develop, research, deploy, or operate agents, this is your community.

Membership is free, and it stays open to anyone who builds, wherever they are.