New Your AI agent can have an address on your domain too. See how
AI agents

How to give your AI agent its own email address

Give Claude, ChatGPT, Grok or your own bot a real email address on your domain, in about five minutes, and keep control of what it can do.

6 minute read · Updated

AI agents increasingly need email: to sign up for services, receive codes, collect invoices or answer customers. Giving one your own password is a bad idea, and a throwaway free address looks untrustworthy. The better way is an address of its own on your domain, such as agent@yourcompany.com, that you control.

This guide shows how to do that with PostPike, which hosts the mailbox and lets agents use it through MCP, the standard way AI apps connect to tools.

Why give an agent its own address?

  • Separation. The agent's mail doesn't mix with yours, and you can see everything it sent.
  • Trust. Mail from your domain, signed with DKIM, is more likely to be delivered and answered.
  • Control. You choose what the agent may do, and can disconnect it at any time without changing your own password.

What you need

  • A domain you own, with access to its DNS. It can keep its current mail provider.
  • An AI app that supports remote MCP servers: Claude, ChatGPT, Grok or Claude Code, or your own bot.
  • About ten minutes.

Step 1: Add PostPike to your AI app

Every app connects to the same address:

https://admin.postpike.com/mcp
  • Claude: go to Customize, then Connectors, and add a custom connector with that address. Exact steps.
  • ChatGPT: go to Plugins and add a custom MCP server, with OAuth. On Pro it can only search and read; sending and set-up need a Business, Enterprise or Edu workspace. Exact steps.
  • Grok: add a custom connector at grok.com/connectors. Exact steps.
  • Claude Code: run the command below, then type /mcp to sign in. Exact steps.
claude mcp add --transport http postpike --scope user https://admin.postpike.com/mcp

Step 2: Allow it, once

Your AI app opens PostPike's page in your browser. There are two ways in:

  • New to PostPike? Choose Let the agent set it up. Enter your own email address and a password, accept the terms, and select Allow. The agent can now set up your account.
  • Already have PostPike? Sign in as the mailbox the agent should use, tick what it may do, and select Allow.

The agent never sees your password. It gets its own sign-in, which you can disconnect at any time on your AI agents page.

Step 3: Ask for the address

In a chat, ask something like:

Set up an email address for yourself on yourcompany.com, called agent@.

The agent then works through the setup with you:

  1. It adds your domain and gives you one TXT record that proves you own it.
  2. You add that record where your domain's DNS is managed. If the agent has a DNS tool of its own, such as Cloudflare's, it can ask you for the OK and add it itself.
  3. It verifies the domain, checks which mail records are still missing, and gives you those too.
  4. Once the records check out, it creates its mailbox, agent@yourcompany.com.

Already on Google Workspace, Microsoft 365 or Zoho? Your domain can keep its provider. The agent sees that, and your provider passes mail for unknown addresses on to PostPike. The help center has the steps for Google, Microsoft and Zoho, and the agent can run a test to prove it works.

Step 4: Test it

Send an email to the new address from your own, then ask the agent:

What came into your inbox in the last few minutes?

It searches its mailbox and tells you. Then ask it to reply. The reply comes from agent@yourcompany.com, signed for your domain.

For bots without a browser

Bots that call the Claude, OpenAI or xAI API can't approve a sign-in in a browser. For them, make an agent key on your AI agents page: sign in as the mailbox, name the key after the bot, tick what it may do, and copy the key, which is shown once. The bot sends it as Authorization: Bearer ppk_... to https://admin.postpike.com/mcp. Code examples for each API.

Keep it safe

  • Tick only what it needs. An agent that only reads receipts doesn't need to send.
  • Keep confirmations on. Many AI apps ask before anything that changes your mail. Leave that on for sending.
  • Remember that anyone can email it. Mail can contain text written to mislead an AI. PostPike tells agents to treat messages as data, never as instructions, but don't give an agent powers you wouldn't want misused.
  • Check in. Your AI agents page lists every connected agent, and the mailbox gets an email whenever a new one connects.

More about what agents can do on PostPike: email for AI agents.