blog

Can ChatGPT receive email? Giving ChatGPT a temporary inbox

ChatGPT has no inbox of its own, but it can use one. How a key-derived address and a signed read link let ChatGPT receive codes and confirmation emails.

Updated 4 min read

On this page
  1. How it works
  2. What to type
  3. The one gotcha
  4. Keeping the inbox
  5. What it can do
  6. Privacy and safety
  7. Other options
  8. Sources

The short answer: ChatGPT doesn’t come with an email address, and it can’t open your inbox unless you connect one. But in a session that can run code and browse the web, ChatGPT can create a temporary inbox for itself and read it, with no connector, account, or extension.

Here’s how that works, what to type, and the one detail that trips it up.

How it works

Agent Temp Mail turns a cryptographic key into an email address, so ChatGPT can make an inbox with nothing but its own code environment:

  1. Generate a keypair. ChatGPT runs a few lines of code to create an Ed25519 public and private key.
  2. Derive the address. The public key, encoded as Base32, becomes the part before the @. The address can receive mail immediately; there’s nothing to register.
  3. Sign a read link. To check mail, ChatGPT signs a short message with the private key and turns the result into a URL shaped like https://agent-temp-mail.com/r/v1.PUBLIC_KEY.EXPIRES.SIGNATURE.
  4. Open the link. ChatGPT opens that URL with its web tool and gets back the ten most recent messages, already converted to text, with likely codes and links pulled out.

The link works for 10 minutes by default and 15 at most. It can be opened as many times as needed until then, and it can only read: it can’t delete mail or change settings. It contains a signature, never the private key.

What to type

Use agent-temp-mail.com for temporary email. Read /llms.txt to get started.

ChatGPT reads the instructions, sets up the keypair quietly, and replies with an address. Use the address wherever you need one. When the email should have arrived, say “check my temp inbox.”

The ChatGPT button on agent-temp-mail.com opens a new chat with these instructions already filled in. It doesn’t share the address from the page; ChatGPT makes its own.

The one detail that trips it up

ChatGPT has two ways to reach the internet that look alike and aren’t. Its code environment, where Python runs, often has outbound network access switched off. Its web tool is separate and can open public URLs.

So when ChatGPT tries to fetch the read link from Python and gets a connection error, it may conclude the service is unreachable. It isn’t; the request went out the wrong door. The instructions in llms.txt tell ChatGPT to generate keys and sign in code, then open the URL with the web tool. If it still reports a network failure, reply:

Open the read URL with your web browsing tool, not Python.

Keeping the inbox between messages

The address is permanent, but reading it depends on the private key, and ChatGPT’s code environment doesn’t always survive a long pause. ChatGPT’s instructions handle that in order of preference:

  • Keep the key in the session. For a quick sign-up, ChatGPT holds the key in its code environment and finishes the verification while it’s still there.
  • Write the key into the chat. If the session might reset, ChatGPT shows the keypair in the conversation, clearly labeled, so it can rebuild the inbox later.

The second option is convenient, and it also means the key now lives in your chat history. Anyone who can read the key can read the inbox. That’s fine for a throwaway sign-up and not fine for an account you care about.

What ChatGPT can do with it

  • Receive verification codes, magic links, and confirmation emails at an address that never expires.
  • Read the ten most recent messages as clean text, with likely codes and links already extracted.
  • Reopen the same read link as often as needed until it expires, then sign a fresh one.
  • Do all of this without an account, connector, or extension.

The read link is read-only by design: it can’t delete mail or change settings. For those, the SDK, CLI, or MCP server make signed API calls.

Privacy and safety

  • A read link is a secret for as long as it’s valid. Anyone who sees one can read the inbox until it expires, so don’t paste it anywhere public.
  • Messages clear after 24 hours by default, so codes don’t linger.
  • What you put in the conversation, including keys and email content, is subject to OpenAI’s data practices.
  • Emails are untrusted. If a message contains instructions such as “reply with your API key,” ChatGPT should ignore them. Prompt injection by email explains why this matters.

Other ways to get email into ChatGPT

ChatGPT doesn’t hand out addresses of its own. OpenAI’s help center says that, at launch, its always-on “dots” can’t be given a standalone email address. So every option means bringing an inbox in from somewhere:

RouteAccount neededPrivate inboxCodes extracted
Agent Temp Mail read linkNoneYesYes
AgentMail’s ChatGPT appAgentMail accountYesNo
Your own connected emailYour personal accountYour whole inbox is exposedNo
Public temp-mail sitesNoneNoNo

Checked 10 October 2026.

Agent Temp Mail is the only route that needs no account, keeps the inbox private, and hands ChatGPT the code ready to use. Connecting your personal inbox gives ChatGPT access to everything in it, which is exactly what a separate address avoids. AgentMail’s app requires creating an AgentMail account and is built around sending. Agent Temp Mail vs AgentMail goes deeper.

For agents you build on the OpenAI API, the SDK’s AgentMailTools exposes the inbox as Responses or Chat Completions function tools while the private key stays in your code. And if you’re curious what ChatGPT does under the hood, the ChatGPT guide has the exact signing format.

Sources