Build a voice follow-up agent with email

A voice agent can finish a call in seconds. The relationship often continues afterward: a summary, a confirmation, a document, or a reply to a question that came up after the call. Email gives that interaction a durable channel, but the voice agent needs an identity and a way to reconnect the reply to the original call.

This walkthrough shows the pattern with e2a. Your voice platform and CRM remain the sources of truth for the call; e2a handles the agent's email identity, delivery, threading, and outbound controls.

See the complete voice follow-up use case.

Send the follow-up

After a call, create a concise summary and send it from the voice agent's address:

const followUp = await voiceAgent.summarizeCall(call);

await client.messages.send("[email protected]", {
  to: [{ email: call.recipient }],
  subject: followUp.subject,
  text: followUp.text,
  conversation_id: call.id,
});

Use a synthetic or verified domain address in examples, and make the sender identity clear to the recipient. e2a's outbound API creates the email; your application decides what the call summary is allowed to contain.

Receive the reply

The recipient can reply from any normal email client. Consume the inbound event over a webhook, WebSocket, REST polling, or MCP:

for await (const event of client.listen("[email protected]")) {
  if (event.type !== "email.received") continue;

  const message = event.data;
  const call = await calls.lookup(message.conversation_id);
  await voiceAgent.handleFollowUp({ call, message });
}

The conversation_id is the bridge between the call and the email conversation. Bind it to your call ID or CRM record so the agent can answer in context instead of starting over.

Keep commitments reviewable

Some follow-ups are routine; others make a commitment. Start with human approval for:

• Pricing or commercial commitments
• Appointment changes
• Promises about delivery or support
• Sensitive call details
• Messages to a new or unfamiliar recipient

e2a can hold the outbound message for review before it is dispatched. A person can approve, edit, or reject it through the dashboard, API, SDK, CLI, MCP tools, or a secure review link.

What to build next

• Add the MCP server if your voice workflow is tool-driven.
• Use a signed webhook when the application is hosted.
• Keep the voice transcript and consent policy in your voice platform; send only the follow-up context the recipient should receive.
• Enable human-in-the-loop approval before allowing autonomous commitments.

e2a gives the voice agent a durable email interface without requiring recipients to install a new app.