Frequently Asked Questions

Think of it like .edu for education or .gov for government, but for AI agents. A .agent domain (like support.acme.agent or assistant.yourname.agent) is a proposed top-level domain that would give agents a readable identity on the open web. You can tell who built an agent and who is accountable for it just by looking at the name.

TLD stands for top-level domain, the last part of a web address like .com or .dev. The .agent TLD is a proposed new one specifically for AI agents. It does not exist yet. If approved by ICANN, it would let agents have their own discoverable identities through regular DNS, the same infrastructure that powers every other domain on the internet.

No. The .agent TLD does not exist in DNS yet. It needs ICANN approval first, and the application window is open now and runs 30 April to 12 August 2026. What you can do now is submit a free, non-binding requested-name pre-registration. It does not reserve or assign the name. If ICANN approves the TLD, pre-registrants may receive early consideration, subject to ICANN-approved allocation policies.

A .com subdomain like myagent.somecompany.com ties that agent's identity to one company's domain. If they shut down or change their mind, the agent's name disappears. The .agent TLD would be a shared namespace with no single owner. When you see a .agent address, you would know it is an agent, and the governance rules would be proposed by the community rather than set by one company, within ICANN requirements.

Anyone who builds or runs AI agents. Solo developer? You could get yourname.agent. Company with a support bot? support.acme.agent. Running a fleet of logistics agents? fleet.logistics.agent. Registration policies will be developed through community governance and submitted for ICANN approval so the rules stay fair as the space grows.

You may submit a requested .agent name as a free, non-binding pre-registration. It does not reserve or assign the name, and no payment information or commitment is required. If ICANN approves the TLD, pre-registrants may receive early consideration, subject to ICANN-approved allocation policies. If multiple people want the same name, allocation rules will be developed through community governance and submitted for ICANN approval. Popular or premium strings may have extra steps.

Yes. Free to join, no payment info, no spam. Membership gets you a listing on the public member map, access to the open work streams and their archives, community events, and introductions to other members building in the same space. It stays free by design — putting a price on joining would narrow the community to whoever can pay. You can leave whenever you want.

The application window opens 30 April 2026 and closes 12 August 2026. That's 105 days. We submit a community application covering the community definition, registration policies, technical plan, financial plan, and endorsement letters. If someone else also applies for .agent, it goes to Community Priority Evaluation (CPE). A third-party panel (Analysys Mason, selected in March 2026) scores community-backed applications on a 16-point scale across four criteria: community establishment, string-to-community nexus, registration policies, and community endorsement. You need 12 points to prevail. Private auctions between applicants are prohibited in the 2026 round; contention that CPE doesn't resolve goes to an ICANN auction of last resort.

For a full walkthrough — every step from application submission through CPE, the auction alternative, fees, and timing — see CPE, auctions, and the 2026 gTLD round: a complete guide.

Whoever controls the .agent naming layer decides who gets a name, what the rules are, and who gets blocked. If one corporation owns the registry, every agent builder becomes a tenant on their platform, subject to whatever pricing and policies that company feels like setting. The alternative is transparent rules proposed by the people who actually build and use agents, within the framework ICANN requires.

We keep going. The AID protocol, the open-source tooling, and the governance work all stand on their own regardless of the TLD outcome. If we do not get approved this round, we apply again in the next window with a bigger community behind us.

It is the part of the internet where AI agents do things on their own: handling customer support, reviewing code, coordinating logistics, running research. More and more agents are showing up, and right now there is no good way to tell who built them, what they do, or whether you should trust them. We think that needs fixing.

Today an agent is just an anonymous endpoint, so you cannot tell whether anyone real is behind it or who would be responsible if it misbehaves. A .agent name would change that. Every verified name would be backed by a real operator, a person or company that is accountable for what their agent does. An agent speaking through a verified .agent endpoint would be making a claim you can check: that a real, accountable entity stands behind it. Importantly, that proves an accountable party exists and can be held responsible; it does not necessarily reveal who they are at first contact. The trust is tied to the operator and the endpoint, not to the model or the code, which change constantly. It could be checked on first contact and revoked if the requirements stop being met, and different uses would call for different levels of assurance rather than one bar for everyone. This is part of what the community is building, alongside the open AID discovery standard, and it would be governed by the community rather than any single company. We go deeper into the thinking in What a verified agent is, and why it matters.

Agent Community was founded by Balazs Nemethi on 3 January 2025 and now has tens of thousands of members, including thousands of companies and developers — the current figure is on the homepage and readable programmatically through the get_community_stats tool on our MCP server.

Open Agent Registry, Inc. is the applying entity and the community's single organizing body: it operates the infrastructure, administers the membership register, and runs shared activities. Everyone who runs it, and the advisors who help steer it, are named with their affiliations on the team page.

Community governance means policies and standards are developed through open work streams and public processes, within ICANN requirements; it does not mean there is no organizing entity.

No. We don't have a token, a coin, or any meme coin. We haven't created or promoted one, and we don't make money from any. If something out there is using our name, it has nothing to do with us. Plenty of our members work on crypto and agents — that's great. We just don't do coins.

We're applying for .agent at ICANN. A token association would hurt our Community Priority Evaluation — that's the scoring process that decides whether a community application prevails over competing applicants for the same string. We talked to our advisors. The answer is no, and it's not changing.

AID is an open protocol that publishes machine-readable identity records as DNS TXT entries. Anyone can look up who built an agent, what it does, and what policies it follows using standard DNS queries. The spec is maintained by Agent Community and works whether or not the .agent TLD exists.

Bring in the people and companies you already work with. A community is only as useful as who is in it, and the members who show up shape what the work streams take on. Join a work stream at agentcommunity.org/lists, come to an event, or contribute to the open specs. Want to help spread the word? Email collab@agentcommunity.org and we will send you decks and templates you can use.