Versine vs. magic links for AI agents
A magic link proves that whoever typed an email address can open that inbox. For a person, that’s a convenient login. For an AI agent signing up on someone’s behalf, it’s a wall, or a hole.
Last reviewed September 25, 2026
What happens when an agent meets a magic link
- The agent reaches your signup page and types its user’s email address.
- Your platform sends a link or a one-time code to that inbox.
- The agent can’t open it. It stops and asks its user to check their email, and the task that brought it to your platform waits until they do.
From your side, that looks like an abandoned signup. From the user’s side, the agent they delegated the task to just handed it back.
Most agents don’t have an inbox, and shouldn’t need one
Browser agents operate a browser. Email is a separate integration that the user has to connect deliberately, and many never do. Some agents run in cloud browsers with no access to the user’s accounts at all. Workplace policies often forbid connecting a work mailbox to a third-party agent.
Even where it’s possible, inbox access is a large grant for a small task. A mailbox holds password resets and login codes for nearly every other service the person uses. Asking users to hand that to an agent so it can click one link on your site is a poor trade, and one careful users will refuse.
Inboxes built for agents invite spam and abuse
The other answer is to give agents inboxes of their own. Services now create email addresses through an API in milliseconds. That solves the agent’s problem and removes yours: if a verified email address costs nothing, verifying one no longer tells you anything.
Signup is already the most attacked door on most platforms. Auth0 reports that 46% of daily signup attempts on its platform in 2024 met its attack criteria. Cheap, programmatic inboxes make that door easier to push on.
- Unlimited “verified” accounts. Every free trial, credit grant or referral bonus gated on email verification can be claimed again with a new programmatic inbox.
- No one to hand the account to. An account tied to an agent’s throwaway address has no path to the person the agent works for. When that person wants to log in, you have no link between them and the account.
- Email as an attack surface. A signup form that sends mail to any address an automated visitor types can be used to flood other people’s inboxes, which hurts your sending reputation and deliverability for real users.
How Versine handles the same signup
Your signup page gets an “Are you an agent?” button next to your existing login options. What happens next depends on the agent.
Agents without Versine
The agent creates a claimable account on its own. Its browser solves a short, automatic proof-of-work check, then your platform receives an ordinary OpenID Connect login for a restricted account that lasts 24 hours. No inbox, no human, no waiting.
The account’s email address is a generated identifier that can’t receive mail, and Versine marks it email_verified: false. Signed claims tell your backend the account is unclaimed and what it’s allowed to do, so it can’t pass as a verified user.
The agent gives its user a private claim link. When the user opens it, they sign up for Versine, verify their real email and accept your terms. Your platform keeps the same user identifier, so everything the agent set up is already there.
Agents connected to Versine
If the user has already connected their agent to Versine, it signs up for a full account from the start, with the user’s verified identity. The user approves the signup on their phone instead of digging through their inbox.
Cost instead of inboxes
Every claimable account costs a fresh proof of work bound to a single signup, and that cost rises for sessions that keep creating accounts nobody claims. Agents that do real work for real people stay at the baseline. Farming accounts gets slower with every attempt.
Side by side
| Dimension | Magic links | Versine |
|---|---|---|
| Needs an inbox | Yes, for every signup and login | No. The person verifies their email once, when they claim |
| Agent without email access | Stalls until the user opens the email | Signs up autonomously with a claimable account |
| Agent with the user’s identity | Needs broad access to the user’s mailbox | Uses the user’s Versine identity, with approval on their phone |
| Cost of creating many accounts | One programmatic inbox each | A proof of work each, rising for unclaimed accounts |
| What your backend knows | An address was reachable | Signed claims: agent signup, account state, deadline, permissions |
| Handing the account to a person | No defined path | Private claim link, same user identifier |
| Email triggered by agent signups | One per attempt, to any address typed | None |
Keep your magic links
Magic links are a good login for people, and Versine doesn’t replace them. It connects to your existing auth provider through OpenID Connect and sits next to the options you already offer. People keep signing in the way they do today. Agents get a way in that doesn’t depend on an inbox.
Frequently asked questions
Can’t an agent just ask its user to click the link?
It can, and today that’s usually what happens. It means the person has to be present for every signup and many logins, which defeats the reason they delegated the task.
Why not accept agents that have their own inboxes?
Because an inbox created by an API proves only that someone called that API. Accepting those addresses as verified makes every email-gated limit on your platform as cheap to bypass as creating an inbox.
Does the claimable account’s email receive anything?
No. It’s a unique identifier for your records, clearly marked as unverified. The person’s real, verified email reaches your platform after they claim the account and sign in again.
Do I need to change my auth provider?
No. Add Versine as an OpenID Connect connection in the provider you already use. The integration guide walks through it.