Require strong sign-in¶
Owners and administrators can open Settings → Security → Account sign-in
policy and require strong sign-in for Owners and admins or All members.
The default is Off. Administrators are the primary owner, platform superusers,
system owners and holders of manage_account, including roles inherited from
teams. A custom role name alone does not grant administrative access.
An authenticator app (including its recovery codes), a user-verified passkey, or
SSO satisfies the policy. Configure MFA requirements inside your identity provider
separately; Preloop accepts its successful SSO authentication.
Enabling enforcement requires the administrator's own strong session. The change ends covered members' non-SSO sessions and returns a replacement session for the administrator. Enforcement is immediate: there is no grace period or scheduled deadline. Tell members before enabling it and check coverage first.
Self-hosted walk-through: Turn on two-factor sign-in for a self-hosted team.
A covered member who signs in without a strong factor is sent to Protect your sign-in. That session can enroll an authenticator app or passkey, refresh, sign out, or switch to an account that permits it. Ordinary APIs and WebSockets refuse it. Confirming the authenticator code restores access after the recovery codes are saved; a new passkey must be used to sign in. Weak sessions cannot switch into a covered account. Refresh rechecks the current policy and role.
Personal API keys, runner credentials and CI identities retain their separate credential policies. Single-user open-source installations remain unchanged.
Coverage and recovery¶
The coverage table pages through members (50 per page); summary counts cover the
whole account. The coverage API accepts offset and limit (maximum 200).
members and members_pending describe that page, while
members_pending_total counts all pending members.
The coverage table shows each member's passkey, confirmed authenticator app and
enforced SSO status. The summary partitions members into enforced SSO, local
factor, and no-factor groups; the authenticator count is a subset of the local
factor group. Unconfirmed enrollment does not count. A required last factor
cannot be removed (mfa_required_by_policy).
Reset sign-in requires a reason and the administrator's own strong proof within ten minutes (or the account window, if shorter). Verify the person's identity independently first. Reset deletes passkeys, authenticator factors and recovery codes, clears passkey-only settings, and ends sessions. The member must sign in and enroll again. The administrator cannot reset themselves, including through another membership of the same person.
A reset is allowed only for a person who belongs solely to this account, or whose email domain is verified for the account. Passkeys are shared across memberships; for a managed shared identity the reset clears local factors and ends sessions across all memberships. The table hides reset for unmanaged identities. Suspended or quarantined identities return a conflict and must be restored before a console reset.
Policy changes write account.sign_in_policy.updated with actor and old/new
settings. Resets write account.member.sign_in_reset with actor, target, reason
and the administrator's authentication method/time. Factor secrets are never
included. Reading coverage writes no audit event.
Step-up window¶
Choose 0 to disable the account window, or 5–720 minutes to require a recent strong proof for covered members before console approval decisions, API-key creation/deletion, SSO and SCIM configuration writes, sign-in policy writes and role changes. The console requests a code or passkey proof and retries the action once. Old SSO sessions return to the identity provider.
Mobile, watch, Slack and Mattermost approval channels do not require console step-up. They use device pairing and channel identity. The chat broker's internal action token is restricted to its signed method and path; adding a channel header to a console request does not bypass step-up.
The window can be enabled only with an active sign-in policy. Turning the policy
off in the console also sets the window to zero; the API rejects off with a
nonzero window (strong_sign_in_policy_required). Member resets always require recent admin proof,
even when this optional window is off. Platform-admin impersonation separately
requires strong admin authentication within ten minutes and records that proof
in the impersonation audit. Resuming the admin session preserves the proof's
original time; it does not extend the freshness window. Impersonation keeps the
admin proof separate from the target's sign-in method, including for passkey-only
targets. It cannot be exchanged for another session or account.
Last-administrator recovery¶
No console session is required for server-side recovery. After verifying the operator and affected administrator independently, run from the Enterprise installation with its configured database:
python scripts/ee/reset_sign_in_policy.py --account <account-uuid> --reason "Lost device; identity verified"
This disables the policy and window and writes an audited server-side policy change. It exits nonzero for an unknown account. Then reset the lost personal factor from the core installation:
cd preloop
python scripts/reset_mfa.py --email admin@example.com --reason "Lost authenticator; identity verified"
python scripts/reset_passkey_only.py --email admin@example.com --reason "Lost passkey; identity verified"
Use the personal reset command(s) matching the lost factor. Sign in, enroll a new factor, save recovery codes, and re-enable the policy. Cloud support uses the same operator recovery path. Do not put a secret or recovery code in an audit reason.
Installation and rollback¶
Deploy the companion core migration, backend and frontend together with this
plugin to enable the controls and enrollment screen. The plugin detects both the
Account.strong_sign_in_policy field and register_sign_in_strength_policy
hook. On an older pinned core it advertises the capability as false, does not
mount policy routes or register enforcement, and leaves existing SSO available.
The companion core also forwards bounded impersonator proof to both the SSO gate
and the strength hook. On the pinned core, only existing same-account SSO
impersonation keeps its original connection context; foreign SSO connections are
never copied.
The social-sign-in second-step screen also requires the companion frontend. CLI MFA continuation additionally requires the companion core callback-binding helper; older cores safely fall back to browser sign-in without forwarding CLI tokens.
To roll back enforcement, set the policy to Off through the console, API or server script. Removing the plugin leaves the core column and hooks inert. Back up before downgrading migrations. Disabling TOTP globally does not disable this policy: covered password sessions still go to enrollment; keep passkey or SSO available, or use the server policy reset during recovery.