> ## Documentation Index
> Fetch the complete documentation index at: https://docs.commune.email/llms.txt
> Use this file to discover all available pages before exploring further.

# Do AI agents need their own email infrastructure?

> Why shared ESP accounts break for autonomous agents and what agent-native email provides.

## The short answer

Yes. Calling a generic email API from a shared account is not the same as having email infrastructure. The difference becomes clear when you need replies, reputation isolation, reply routing, or scale.

## What breaks with shared email infrastructure

**Calling `resend.emails.send()` or `sendgrid.send()` from your agent gives you one thing: outbound delivery.**

It does not give you:

| What you need | What shared ESPs give you |
| - | - |
| An inbox that receives replies | ✗ No |
| Reply routing back to your agent | ✗ No |
| Per-agent sender reputation | ✗ Shared reputation for all sends |
| Automatic thread resolution | ✗ No |
| Programmatic inbox creation | ✗ No (dashboard only) |
| Warmup per agent/campaign | ✗ Manual, one domain |
| Bounce/complaint rates per agent | ✗ Account-wide only |

## Why replies matter

Email is a conversation. Your agent sends an outreach email. The prospect replies. That reply needs to go somewhere.

With a shared ESP, that reply either:

* Goes to a mailbox no one checks
* Goes to a human inbox the agent can't read
* Gets lost

With Commune, the reply is delivered to your webhook with the full parsed email, threaded correctly to the original send.

## Why reputation isolation matters

If your agent sends to a bad list, your complaint rate spikes. With shared infrastructure, that spike affects every sender using the same account.

With per-agent inboxes, one bad campaign doesn't poison your other agents' sending.

## Why programmatic creation matters

Your agent can't open a browser, navigate a dashboard, fill out a form, and create an inbox. It needs to call an API.

```typescript theme={null}
// This is what agent-native email looks like
const inbox = await commune.inboxes.create({ localPart: 'outreach-q1' });
// → outreach-q1@commune.email, ready immediately
```

## What "agent-native" email infrastructure means

1. **Programmatic inbox creation** — one API call, no dashboard
2. **Send + receive** — not just outbound
3. **Automatic threading** — replies resolve to the correct conversation
4. **Per-inbox reputation** — isolation between agents and campaigns
5. **API-accessible metrics** — bounce rate, complaint rate, suppression list
6. **Automatic warmup** — domain reputation built gradually

## Related

<Columns cols={2}>
  <Card title="How do I give my AI agent an email address?" icon="inbox" href="/kb/how-do-i-give-my-agent-an-email-address">
    Create an agent inbox with one API call — no domain setup required.
  </Card>

  <Card title="How do I handle email replies in my agent?" icon="bolt" href="/kb/how-to-handle-email-replies">
    Configure webhooks to receive inbound emails and route replies back to your agent.
  </Card>

  <Card title="How do I set up per-agent inboxes?" icon="inbox" href="/kb/per-agent-inboxes-multi-agent">
    Create isolated inboxes per agent or campaign for reputation and reply routing.
  </Card>
</Columns>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.