USE CASE · SUPPORT

Build an AI support agent that works from a real inbox

e2a lets developers build support agents that receive customer email at a real address, inspect SPF/DKIM/DMARC authentication evidence, draft or send a threaded reply, and route sensitive outbound messages through human approval.

Build your support agent →See the runnable example ↗

THE WORKFLOW

What the agent does

  1. 01A customer emails your support address.
  2. 02e2a receives the message and evaluates sender authentication.
  3. 03Your agent reads the structured event and retrieves application context.
  4. 04The agent classifies the request and drafts a response.
  5. 05Low-risk replies send automatically; sensitive replies wait for approval.
  6. 06The response stays in the original email conversation.

WHY E2A

  • ✓SPF, DKIM, and DMARC evidence on inbound mail
  • ✓Signed webhooks, WebSockets, REST polling, or MCP delivery
  • ✓Conversation correlation and threaded replies
  • ✓Per-agent human-in-the-loop approval
  • ✓Optional inbound prompt-injection and phishing screening

START BUILDING

The integration stays small

Support agent runbook ↗
async for event in client.listen("[email protected]"):
    if event.type != "email.received":
        continue

    message = event.data
    reply = await support_agent.run(
        message["conversation_id"],
        message["text"],
    )
    await client.messages.reply(
        message["delivered_to"],
        message["message_id"],
        {"text": reply},
    )

The agent owns the business logic. e2a handles the email identity, delivery, authentication evidence, and reply routing.

KEEP GOING

Connect the pieces

Quickstart →
Create an inbox and send your first message.
MCP server →
Give a local or hosted agent an inbox without an API key.
API reference →
Explore the stable /v1 operations.
Blog tutorials →
Follow framework-specific integration guides.

FAQ

Can an AI support agent receive email without a public URL?

Yes. An e2a support agent can receive inbound email over a WebSocket stream, REST polling, or MCP, so a local or private application does not need to expose a public webhook endpoint.

How does e2a verify a support request's sender?

e2a evaluates SPF, every DKIM signature, and DMARC on inbound mail, then delivers the authentication results as structured evidence for the support agent to use in its decision-making.

Can a human approve an AI-generated support reply?

Yes. Human-in-the-loop approval can hold outbound messages for review before they are sent, with approval available through the dashboard, API, MCP tools, CLI, or a magic link.

Does e2a preserve email threads?

Yes. e2a supports reply-in-thread delivery and exposes a stable conversation_id so the application can connect an email conversation to the support agent's own session state.

Build your support agent

Open-source, free to start, and ready for your first agent workflow.

Get started →