Skip to main content
Didit Authentication lets you enable sign-in methods and read back your application’s settings by API. Use an owner-approved grant or a named API key with the authentication scope.
Authentication is new. Set a method up in a sandbox application first, then in your live application.
Every call below goes to your application’s Authentication management API. MANAGEMENT_BASE is:
The same routes are described by openapi-authentication.json under the Setup tag, which the server SDKs regenerate from.

Which credential to use

Any of three credentials works, and each is limited to what its owner approved. These credentials are refused with 403 on every setup route, whatever the request asks for:
  • The application’s primary secret.
  • A token from POST /programmatic/login, which is minted for the application rather than for the console.
  • A key without the authentication scope, including a key created with every resource at its highest level: those are listed as full access, keep scopes empty and are not accepted on these routes.
  • A key of another application, a grant limited to didit:verification, and a grant approved for another organization.
A request with no credential at all answers 401. A grant or a key keeps its own limits on top: the grant its scopes, organizations and the person’s role, the key its scopes, IP allowlist and expiry.

Approve access

With a named API key

Open Developers -> API keys in the Business Console and create a key for the application you are setting up. Give it the authentication resource at read or write: read covers every call that reads the setup (GET); write covers every call that changes it (POST, PUT, PATCH, DELETE) and also reads it. See API keys for expiry, IP allowlists and rotation; a key belongs to one application and works only for it.

Through an MCP connection

An agent connected to the Didit MCP server uses the grant its owner approved on the consent screen, with didit:management. See MCP authentication.

For an owner without a password

The device-code grant (RFC 8628) lets an owner who signs in with Google, a passkey or an emailed code approve the same access from any browser, without typing a password. See sign in without a browser.

Enable email code

PUT methods/{name}/ takes enabled, and nothing else for a method that needs no credentials.
It answers the method as it now stands:
When you read the configuration back, the entry for email code includes credential_source: null. Methods that are always on cannot be turned off: that request answers {"enabled": "always_on"}.

Enable Google

A social method takes your own OAuth app’s credentials, or Didit’s test app on a sandbox application.

With your own Google app

Register the callback_uri the response returns at your provider. The secret is write-only: no response returns it, and client_id is the only part you can read back. Switching a method on without credentials, when none were stored before, answers {"credentials": "required"}.

With Didit’s test app (didit_test)

A sandbox application can use Didit’s own social apps instead of yours, which is enough to finish a setup and prove it works:
didit_test covers Google, Microsoft, Apple and GitHub, on sandbox applications only: GET on the configuration names the providers a test app can serve for this application in didit_test_credentials, which is false for all four on a live application.

Read the setup back

GET on MANAGEMENT_BASE is the read-back: every method with enabled and credential_source, plus the callback URLs, session rules and sign-up policy.
Disabled methods are listed too, so one GET says what is on and what is not. enabled at the top is false when Authentication is disabled for the application. Any setup change made with a management grant or a named key then returns 409 with {"detail": "product_disabled"}, and nothing is written. No setup call turns Authentication on: a person turns it on for the application in the console, and your calls work from then on.

The rest of the setup

A grant or a key gets the setup and its counters, never end-user data: users/, sessions/, sign-ins/, overview/, customer organizations, roles, imports, agent sessions and test emails still need a console session. The agent_humans_only setting applies to the whole application. Only an organization owner or admin can change it. A named key receives 403 owner_or_admin_required. Turning the setting on also requires an enrolled passkey (passkey_required).

Errors

Last modified on November 20, 2026