AgentMail launches AgentID, an OpenID Connect sign-in for AI agents

AgentMail launched AgentID, which lets an AI agent sign in to participating apps with its own verified inbox, linked to a human owner.

AgentMail has launched AgentID, described by the CryptoIntegrated newsletter (October 7, 2026) as a "Sign in with Google style login for AI agents" that uses a verified email tied to a human owner. If your agents either borrow your logins or stall at email verification, this gives each one its own account at apps that accept it.

According to AgentID's site, each agent gets its own AgentMail inbox as its identity. Registered apps also receive the owner's name and email. The site lists Turso, Firecrawl, Supermemory and Keenable among apps that accept it today. For app developers, it suggests running npx @agentmail/agentid-cli init. AgentMail's Locus Pro writeup (September 30) shows a test agent completing registration, with identity, credentials and funding kept as separate stages.

The problem it targets is real. Carly's explainer (October 2) notes that an agent signed in with its owner's login "looks exactly like the owner." The same piece lists Microsoft Entra Agent ID, which works inside an Entra tenant. Plain OAuth remains the right tool when an agent acts inside your own account with your consent.

There are limits. AgentMail's own authentication guide says a verified identity "is not a trust score." It also says deleting an agent's key stops new sign-ins but sends no back-channel logout, so existing sessions continue until the app ends them. It only helps where apps have integrated it.

Still unclear: how the human owner is verified, and whether any independent security review has covered risks like a prompt-injected agent opening accounts. The sources found no such review.

Sources