How monitoring actually scans
You turn it on through the agent or in your settings, and from then on it runs read-only sweeps roughly every half hour against your connected account.
What counts as a match is deliberately narrow: mail from an employer you applied to through Intervieux, mail from the specific hiring platform that application went through, and mail from any extra sender domain you've explicitly told it to watch, like a recruiting agency you're expecting to hear from outside the normal flow.
Anything else in your inbox, personal mail, newsletters, unrelated work, is not part of what this scan touches.
For whatever does match, the AI only extracts the sender, the subject line, and a snippet capped at 500 characters, classifies it, interview invite, offer, rejection, scheduling, and stores that classification so you can search it later with search_correspondence.
You can also trigger check_job_inbox any time for an on-demand scan using that same pipeline, useful right after you know a company said they'd respond soon.
Checking in on demand
It's been a week since you applied to a data analyst role and you want to know if anything came back before you go looking yourself. You ask: "check my inbox now, did any employers reply?"
The agent calls check_job_inbox, a live read-only scan against your connected account using the same targeted matching as the scheduled sweep, and reports back that one message came through, classified as an interview invite, from the company you applied to.
It tells you the sender and the subject line and describes what the snippet suggests, that it looks like a scheduling request, but it doesn't read the message to you word for word, because it never had access to more than that snippet in the first place. From there it can help you draft a reply, which is a separate, approval-gated step, or you can open the actual email yourself to read it in full.
Boundaries
What the agent can never see or do here
The privacy limit is structural, not a setting you configure: even for a message that matches, the AI extracts only the sender, subject, and a snippet capped at 500 characters. It never receives the full body of any email, matched or not, and it never scans mail outside the narrow matching criteria, no blanket read of your inbox under any configuration.
Every scan, scheduled or on-demand, is strictly read-only. It can find and classify a message, but it can never send a reply, delete anything, move anything, or modify anything in your mailbox; that boundary holds even for messages it has already flagged.
Turning monitoring on or off, or adding and removing a watched sender domain, is the one write on this surface, and it requires your approval every time you change it. Monitoring is free, part of the agent, with no subscription required to turn it on.
Frequently asked questions
Does inbox monitoring read the full text of my emails?
No. Even for a message it matches and classifies, the AI only ever extracts the sender address, the subject line, and a snippet capped at 500 characters. The full body of the message is never part of what it reads.
Does monitoring scan my entire inbox?
No. It only matches mail from employers you applied to through Intervieux, the hiring platforms those applications went through, and any sender domains you've explicitly added. Everything else in your inbox is outside its matching criteria.
Can monitoring delete or move emails in my inbox?
No. Every scan, scheduled or on-demand, is strictly read-only. It can find and classify a matching message, but it cannot send, delete, archive, or move anything.
How often does the automatic scan run, and can I turn it off?
Roughly every 30 minutes while it's on. Turning it on or off requires your approval, the same as adding or removing a watched sender domain, and you're notified as interview invites, offers, and rejections come in.
Related pages
Turn on inbox monitoring
Connect Gmail or Outlook and ask the agent to watch for job replies, at no cost. It only ever sees the sender, the subject, and a short snippet, nothing more.