Most AI agent email infrastructure focuses on inbound (receiving OTPs, verification links, and notifications). But a growing class of AI agents needs outbound capability too: agents that nurture leads, send personalized follow-ups, or run automated marketing campaigns on behalf of users.
The Lumbox Outbound Stack
Lumbox provides two outbound primitives:
- Transactional sending:
POST /v1/inboxes/:id/sendfor individual emails from a specific agent address - Bulk campaign sending:
POST /v1/contact-lists/:id/sendfor broadcasting to a contact list
Building an Agent-Driven Campaign
import { Lumbox } from "@lumbox/sdk";
const client = new Lumbox({ apiKey: "ak_your_key" });
// 1. Create or retrieve a contact list
const list = await client.contactLists.create({ name: "Q2 Prospects" });
// 2. Add contacts (bulk import)
await client.contacts.bulkAdd(list.id, [
{ email: "alice@acme.com", name: "Alice", metadata: { company: "Acme" } },
{ email: "bob@corp.io", name: "Bob", metadata: { company: "Corp" } },
]);
// 3. Send the campaign from an agent inbox
const inbox = await client.inboxes.get("inb_abc123");
await client.contactLists.sendCampaign(list.id, {
from_inbox_id: inbox.id,
subject: "Your Q2 infrastructure review",
html: "Hi {{name}}, ...
",
});
Personalization at Scale
The {{name}} and {{metadata.*}} template syntax is resolved per-recipient at send time. An AI agent can generate personalized HTML content per segment, then send each variant to a different contact list, all without human involvement.
Compliance by Default
Lumbox enforces one-click unsubscribe headers (RFC 8058) on all bulk sends automatically. An AI agent cannot accidentally send non-compliant email. Unsubscribe events are tracked and suppressed from future sends, ensuring your contact lists stay clean without manual maintenance.
The Agent Use Case
A typical agent-driven pipeline looks like this: a research agent identifies prospects, an enrichment agent adds company context, a copywriting agent generates personalized email content, and Lumbox delivers it, all in an automated loop with no human in the critical path. The human reviews results and approves the next campaign batch.