Create Agent Inbox

Create and configure hosted Agent Inboxes for AI workflows.

View as Markdown

Pattern 1: Agent with Dedicated Inbox

Each agent gets its own inbox for isolated communication:

const inbox = await client.inboxes.create({
  display_name: "Customer Support Agent",
});

// Agent sends emails from its inbox
await client.inboxes.messages.create(inbox.id, {
  to: ["customer@example.com"],
  subject: "Re: Your request",
  text_body: agentResponse,
  from_: inbox.id,
});

Pattern 2: Webhook-Driven Agent Loop

Use webhooks to create an autonomous email loop:

  1. Agent receives webhook when email arrives
  2. Agent processes the email content
  3. Agent sends a reply
app.post("/webhook", async (req, res) => {
  const { data: message } = req.body;

  // Generate response using your AI model
  const response = await generateResponse(message.text_body);

  // Reply to the sender
  await client.inboxes.messages.replyAll(
    message.inbox_id,
    message.id,
    { text_body: response }
  );

  res.status(200).send("OK");
});

Pattern 3: Semantic Search for Context

Use semantic search to find relevant past conversations:

POST /agent/v0/search
{
  "query": "shipping delay refund policy",
  "limit": 5
}

This returns the most semantically similar messages, which agents can use as context for generating responses.

SDK Compatibility

Agentry is API-compatible with the AgentMail SDK. Point it at the Agentry API:

import AgentMail from "agentmail";

const client = new AgentMail({
  apiKey: process.env.AGENTRY_API_KEY,
  baseUrl: "https://api.agentry.to",
});