MCP server

Disposable email for AI agents

Give Claude Code, Cursor, VS Code or an agent in CI its own inbox. It creates a mailbox, signs up with it, waits for the email and reads the verification code, with no glue code.

Serverhttps://api.fake-email.pro/mcp

Streamable HTTP · signed in with your API key · nothing to install

Sign up on staging.acme.example with a fresh inbox and confirm the account.

fake-email · create_mailbox

{
  "address": "quiet.river.17@example.org",
  "expiresAt": "2026-11-22T09:41:00Z"
}

…fills the sign-up form in the browser…

fake-email · wait_for_message12 s

{
  "subject": "Confirm your Acme account",
  "code": "482913",
  "confirmLink": "https://staging.acme.example/confirm?t=…"
}

Signed up as quiet.river.17@example.org and confirmed the account with code 482913.

Connect your client

One command for Claude Code, one file for Cursor or VS Code, and in CI the key comes from a secret. Once the client has your API key, the agent sees ten new tools.

Full setup, CI and troubleshooting →
claude mcp add --transport http fake-email https://api.fake-email.pro/mcp \
  --header "Authorization: Bearer $FAKE_EMAIL_API_KEY"

Ten tools, the whole API

Everything the REST API does with mailboxes and messages, as tools the agent calls on its own.

create_mailbox
Create a disposable mailbox that receives email for 24 hours.
list_mailboxes
List your active mailboxes.
get_mailbox
Check one mailbox: when it expires and how many messages it has.
delete_mailbox
Delete a mailbox and its messages right away.
wait_for_message
Wait up to 50 seconds for an email and return it with the verification code and confirmation link already extracted.
list_messages
List the messages of a mailbox, newest first.
get_message
Read one message: its text, codes, links and the list of attachments.
delete_message
Delete one message with its attachments.
list_domains
List the domains a new mailbox can use.
get_limits
Show your API quota and rate limits.

Email is untrusted input

Code and link come first

The server returns the extracted verification code and confirmation link first, so the agent rarely needs to read the email body.

The body is fenced off

The email text is wrapped in markers that tell the model not to follow instructions in it. Still, review what your agent does with mail from unknown senders.

No separate price: each mailbox the agent creates counts towards your monthly API quota, just as over REST. Compare the plans on the pricing page or read about the Temp Mail API.

Frequently asked questions

Which clients does it work with?

Any MCP client that speaks Streamable HTTP and can send an Authorization header: Claude Code, Cursor, VS Code, Windsurf and agents you build yourself.

What does it cost?

The MCP server has no price of its own. Each mailbox the agent creates counts towards the monthly API quota of your plan; waiting for an email and reading it take nothing from the quota. A key is free with any account.

Can I run it in CI?

Yes. Commit .mcp.json to the repository and store the key as a CI secret named FAKE_EMAIL_API_KEY: the client reads it from the environment, so the key never ends up in the repository.

Is there something to install?

No. The server runs at api.fake-email.pro/mcp: you give your client its address and your API key, and that is all.

Can my agent safely read email from strangers?

Emails are written by outside senders, so the server hands them to the model as untrusted data: the extracted code and link first, then the email text wrapped in markers that tell the model not to follow instructions in it. Still, review what your agent does with mail from senders you don't know.

Give your agent an inbox

Create an API key, connect your client and let the agent sign up on its own.