Skip to content

Caller verification

Breeze requires caller verification before Microsoft 365 password resets and account disables. A verification is usable once, for the named action and target, by the initiating technician. The default minimum assurance is tier 2 and the default freshness window is 30 minutes.

Open Verify caller on a ticket, contact or device. Confirm the requester, action and target. Choose an available method and explain the action shown on the challenge card. The caller selects the matching number; never ask them to read a code back.

In this release the available methods are SMS link, email link and callback attestation. SMS and email reach tier 2 only through an established destination. Imported or AI-written destinations require technician attestation. Callback attestation is tier 1. Email to any of the target account’s mailbox addresses cannot verify a reset of that mailbox.

Workstation verification is not available in this release. The prompt that appears on a bound workstation (tier 3) ships dark behind CALLER_VERIFICATION_WORKSTATION_ENABLED, which defaults to false. While it is off, the workstation method is reported as unavailable, and tier 3 cannot be selected in the policy: the policy editor and API cap the required tier at 2. Because tier 3 is reachable only through a bound workstation, a policy that required it could never be satisfied.

Number matching confirms the request; it does not prove the technician’s identity or prevent a coached caller from being relayed. If uncertain, the caller should hang up and call the MSP on a number they already trust.

Disable an account without contacting its owner

Section titled “Disable an account without contacting its owner”

An org-level contact with an allowed authoriser role can verify a disable of another account. A site’s primary contact has no automatic org-wide authority. For offboarding or containment, use Administrative disable through the API: enter a reason of at least 20 characters and complete interactive MFA step-up. This option requires organization write permission and an effective policy that permits it. It cannot authorise password resets, and ordinary action approval still applies.

This is not me fences the requester and relevant targets, opens a security incident, and cancels pending or approved actions. An action that already started dispatch appears in the incident for investigation; confirm in Entra whether the change landed. Expired challenges can still report a suspicious request within the rejection window.

A verification remains used when dispatch fails. Re-verify before creating another intent. A retry of the same intent does not consume a second grant, but an already-started dispatch must be reconciled rather than blindly repeated.

Bind contacts to their Entra principals and attest imported destinations before relying on tier 2. Renaming a UPN does not change an intent’s pinned account.

CALLER_VERIFICATION_ENABLED defaults to true. An explicit false hides entry points and refuses protected execution; it does not disable the safety gate while leaving mutations available. Only the exact value true enables the feature, so any other explicit value (including an empty string) turns it off. Self-hosted deployments must map this variable into the API service environment as well as setting it in .env. Deploy the OID-capable Graph actions executor before the API, so pinned actions are accepted when enforcement starts.

CALLER_VERIFICATION_WORKSTATION_ENABLED defaults to false and also requires the exact value true. Leave it unset until workstation verification is announced.