An ops inbox your agent reads first.
Point monitoring alerts, vendor notices and service emails at one address. Your agent reads them as they arrive and hands the ones that need a person to your team.
One address for machine mail.
Create a mailbox on your own domain and give that address to your monitoring, your vendors and your status pages. Everything they send lands in one inbox your agent owns.
POST /v1/agent/mailboxes{ "address": "ops@yourco.com", "displayName": "Ops inbox" }→ { "id": "mbx_3f7a", "address": "ops@yourco.com", "status": "active" }Rules clear the noise.
Mailbox rules run on every new thread before your agent looks. Archive the weekly digests, file a vendor's notices into a folder, and leave the rest for the agent.
molted rules create --mailbox mbx_3f7a \ --name "Archive digests" \ --condition-type subject --condition-operator contains \ --condition-value "weekly digest" --action-type auto_archiveYour agent reads what's left.
It lists the open threads, reads each one, and decides: resolved, worth watching, or a person's problem. Each message comes marked as untrusted content with an injection risk level, so your agent can treat a notice as data, not instructions.
read_inbox({ mailbox_id: "mbx_3f7a", status: "open" })→ { items: [ { id: "thr_a90c", contactEmail: "alerts@uptime.example", subject: "DOWN: api.yourco.com (3 checks)", status: "open" }, { id: "thr_a91d", contactEmail: "billing@cloud.example", subject: "Your payment method expires soon", status: "open" } ], total: 2 }It escalates what needs a human.
When an alert needs someone on call, your agent hands the thread over with what it found. The note says what broke and what it already checked.
handoff_thread({ thread_id: "thr_a90c", note: "api.yourco.com down for 3 checks. Status page is green. Needs on-call." })→ { handoff: { status: "requested" }, created: true }Give your agent its own mailbox.
We'll email you when signups reopen.