Follow one email through the machine.

Every inbound is interrogated before your agent ever sees it. Try it — this demo is the real thing.

intent: "schedule_demo"entities: {  date: "tuesday"  time: "10am"  from_email: "jordan@stripe.com"}urgency: 0.53injection_score: 0.01sender_reputation: 0.91quarantined: false

The email leaves the scanner as data — injection score and sender reputation on every event, intent and entities when you opt in.

{
  "id": "rep_01H8...",
  "agent": "sarah",
  "from": "lead@example.com",
  "subject": "Re: Demo",
  "body": "Tuesday at 10am ET works...",
  "injection_score": 0.02,
  "sender_reputation": 0.91,

  // Only present when classification is enabled on this agent:
  "intent": "schedule_demo",
  "entities": {
    "date": "2026-05-14",
    "time": "10:00",
    "timezone": "America/New_York"
  },
  "urgency": 0.8
}

Six lines, and the wire runs through your function. Same shape in TypeScript, Python, and MCP.

import { mails } from "@mailsai/sdk";

const sarah = mails.agent("sarah", {
  domain: "yourcompany.com",
});

await sarah.send({
  to: "lead@example.com",
  subject: "Demo",
});

sarah.onReply((reply) => {
  // reply.injection_score   → 0.02
  // reply.sender_reputation → 0.91
  // reply.intent   → "schedule_demo"
  // reply.entities → { date, time }
  // You decide what to send next.
});

Every sender is scored; bad actors get paused at 0.3% — before the network ever feels them.

You pay per event, not per month. Quiet months cost zero.