Skip to content

Login methods

An environment’s login methods are the ways its users can prove who they are. Every method is a switch, and the login UI shows exactly the methods that are on, in the embedded login and on the hosted page alike. The apps of a group share one set of methods, because they share their users.

Id Method Way in Second factor
email_code A 6-digit code sent to the user’s address. Yes No
password An email address and a password. Yes No
passkey A passkey made in the user’s account settings. See Passkeys. Yes Yes
google, apple, github, discord, x, microsoft A social login. See Social and custom providers. Yes No
custom-… A provider of your own that speaks OpenID Connect or OAuth 2 (Pro and Scale plans). Yes No
solana, evm A wallet signs a message. See Wallets. Yes No
phone Codes by SMS. Not available yet

The second factors themselves (an authenticator app, a passkey, recovery codes) are the user’s own, set up in account settings, and are described in Two-step login. A passkey works as a second factor as soon as the environment has a passkey domain, even with the passkey login method off.

A new environment starts with email_code on, sign-up open and two-step login optional. Email codes need no setup, so a new app logs users in at once.

In the dashboard, open Login methods for the environment and switch methods on or off. The same with the management API (admin role):

PUT /v1/environments/{environment_id}/login-methods
Authorization: Bearer sk_dev_...
Content-Type: application/json
{ "methods": ["email_code", "password", "google"] }

The set is saved in catalogue order, with custom providers after the built-in ones. A save is refused (400) when:

  • a method is not available yet (phone);
  • passkey is on and the environment has no passkey domain yet;
  • a social method has no credentials in this environment. In a development environment the platform’s shared credentials count; production environments need your own;
  • an id is unknown;
  • no method would stay on.

See Environments for the full request and response.

There are two doors to the same login. The embedded login runs inside your own page (auth-react’s modal, or a UI of your own on /authorize-challenge). The hosted page is served by SeekAuth on the issuer host and is reached by redirecting to /authorize. Both apply the same steps, the same rules and the same limits, and both end with an authorization code for your client. Social and custom providers always leave through a redirect, even from the embedded login. See How login works.

  • A code is 6 digits and expires 10 minutes after it was sent. The email says so.
  • Five wrong answers end a code, whether the sixth would have been right or not. The login UI shows how many tries are left. After that, or once the code expired, the user asks for a new one.
  • Asking again sends a fresh code and the old one stops working. There is no waiting time between codes, but code emails are rate limited per address, per inbox and per caller.
  • The answer is the same whether or not the address has an account. When sign-up is closed, or the account is disabled, the login looks the same and no email is sent.
  • A code to an address nobody has creates the account when it is entered, if sign-up is open. See Sign-ups and access.

Passwords follow length rules only, with no rules about kinds of characters:

Environment’s two-step login Shortest Longest
Optional (the default) 15 characters 64 characters
Required 8 characters 64 characters
  • A new password (at sign-up, at a reset, or from account settings) is checked against known data breaches with the Pwned Passwords range API. Only the first five characters of the password’s SHA-1 hash leave SeekAuth, and the request is padded. A password found there is refused with “This password appeared in a data breach. Choose another one.” If the breach service does not answer, the password is not refused for it.
  • Passwords are stored as argon2id hashes (19 MiB of memory, two passes, one lane).
  • A wrong password, for an account that exists or not, answers “Wrong email or password.” and takes as long either way.
  • Someone whose address is not verified yet is sent a code to verify it first.
  1. The user asks for a reset with their address. The answer is always the same; a code goes only to an account that exists and is active.
  2. They enter the code from the email. A user with a second factor is asked for it now, before anything changes: the code proved the mailbox, which is one factor.
  3. They are asked for a new password, and it is set. Every session of the user ends, with its tokens, they are emailed that the password changed, and the login goes on.

A reset sets a password only for the account that still has the address its code went to: if the account moved to another address while the reset waited, the reset starts again.

The password method has to be on for resets to work. A signed-in user changes or removes their password through the account API.