For hiring teams · Consent-based · Human-reviewed
Proctor runs beside your own interview, in the video tool or coding exercise you already use. The candidate installs a native desktop app and agrees to exactly what is captured; a named reviewer on your team reads the evidence and records the conclusion. It is a separate product from the Luminos Hire AI interview, with its own accounts.
Already a customer? Sign in
macOS, Windows and Linux. No browser extension, no hidden background process.
Both keep a person responsible for the outcome. They differ in who runs the interview. Proctor is a separate product with separate accounts — signing in to Luminos Hire does not sign you in here.
Use it when someone on your team conducts the interview or assessment in your own tools, and you want consented evidence of the session for a reviewer to read afterwards. It records observations; it does not assess skill.
How hiring teams use itUse it when you want an AI-assisted interview session run on Luminos Hire, with its own human-reviewed report and verification tier.
About Luminos Hire interviewsThe two can also work together: a Luminos Hire interview can be linked to a Proctor session so its desktop evidence is available to the same review. The products still keep separate accounts and separate data.
These are not settings. They are properties of how the agent and the API are built, and every release is audited against them.
No capture, event or upload happens until the server holds a consent record naming the exact scopes. Without one, the ingest API refuses the write.
A banner in the app and an indicator in the menu bar or tray, for the whole session. There is no hidden mode to enable — not even for administrators.
No keystrokes, no window titles, no software inventory. Screen, audio and camera cross the network only under a scope the candidate agreed to and your policy allows.
Observations are described, never scored. There is no risk number, no automatic flag and no pass/fail — because the decision belongs to a named human being.
22 capabilities across the agent, the live session, the evidence store and the platform around them — most of which were previously only visible after you logged in.
Swift on macOS, WinUI 3 on Windows, GTK4 on Linux — each using its platform's real capture and permission APIs, and installed as a normal application.
No capture, event or upload happens until the server holds a consent record naming the exact scopes. Without one, the ingest API refuses the write.
Watch a running session from the browser — media, participants and the event stream as it is written.
Timeline, media and transcript in one view, so a reviewer reads a moment and its context together instead of correlating three exports.
Provision sessions, fetch evidence and manage retention from your own systems. The OpenAPI document is published, and the SDK is generated from it rather than hand-written.
Provision the session from your own systems — the API, the Node SDK or an ATS connector. It comes back with a join link per participant, and captures nothing yet.
The candidate installs the desktop app and reads exactly what will be collected. Nothing starts until they agree to it.
Monitoring runs with the indicator visible. If the network drops, the agent buffers locally and resumes — nothing is lost.
A reviewer reads the timeline, media and transcript together, and records their own conclusion in their own words.
Every one of those is a deliberate limit rather than a missing feature. A tool that made the decision for you would move the responsibility away from the person who has to justify it.
Tell us how your team interviews today. A person reads every request and sets up access with you — there is no self-serve sign-up.
Already a customer? Sign in