Email for AI agents
Agents keep running into email: signups want verification codes, tools want confirmed addresses, tests want real delivery. TempMaily gives an agent its own disposable inbox — created, read, and deleted entirely through tool calls, over MCP or the REST API.
One config block, six tools
Paste this into your MCP client — Claude Desktop's claude_desktop_config.json, Claude Code, Cursor's MCP settings — and the agent gains create_inbox, list_inboxes, delete_inbox, list_messages, get_message, and delete_message. The key comes from the dashboard's API Keys section.
{
"mcpServers": {
"tempmaily": {
"url": "https://tempmaily.co/api/mcp",
"headers": {
"Authorization": "Bearer <your API key>"
}
}
}
}From there, "sign up for this service and confirm the email" is a complete agent task: it creates an inbox, uses the address in the form, polls its messages, and pulls the code out of the body it just read. The full walkthrough lives in give your AI agent an inbox.
The same inboxes over REST
Not every pipeline is an MCP client. The REST API exposes the same operations on the same inboxes with the same Bearer key — the right fit for CI jobs and custom frameworks, and the backbone of our email verification testing guide.
# Create an inbox for the agent
curl -X POST https://tempmaily.co/api/v1/inboxes \
-H "Authorization: Bearer $XTM_KEY"
# Poll for the verification email
curl https://tempmaily.co/api/v1/inboxes/<id>/messages \
-H "Authorization: Bearer $XTM_KEY"
# Read the full message (bodyText / bodyHtml)
curl https://tempmaily.co/api/v1/messages/<messageId> \
-H "Authorization: Bearer $XTM_KEY"What agents use it for
- Agents that sign up for things
An agent registering for a service hits the same wall a human does: the verification email. With its own inbox it receives the code and finishes the flow alone.
- QA agents testing email flows
Point an agent at your staging app and it can mint a fresh inbox per run, trigger the signup, and assert the mail actually arrived — the agent version of the E2E patterns in our testing guides.
- Throwaway identities for research tasks
Scraping a gated download or evaluating a tool that demands an email? The agent uses a disposable address, and the follow-up marketing dies with the inbox.
Honest constraints: inboxes are receive-only (an agent can read mail, not send it), programmatic access needs a Premium key, and the usual rate limits apply. If your agent is a person using ChatGPT rather than an autonomous tool-caller, temp mail for ChatGPT and AI tools is the better starting point.
Get a key
One flat Premium plan — $9.90/mo or $90/year — covers the MCP server, the REST API, up to 10 concurrent addresses, and custom domains. No enterprise quote, no usage tiers.
Go Premium
Custom domain, up to 10 addresses, 2GB storage, no ads — from $7.50/mo billed yearly.
Agent email FAQ
Can an AI agent have its own email address?
Yes. An agent connected to TempMaily's MCP server can call create_inbox and get a real, working disposable address back in the tool result. Mail sent to that address is readable by the agent through list_messages and get_message, so the whole loop — sign up somewhere, receive the verification email, extract the code or link — happens inside the agent's normal tool-calling flow with no human checking a mailbox.
What is an MCP email server?
MCP (Model Context Protocol) is the open standard AI clients like Claude Desktop, Claude Code, and Cursor use to give models tools. An MCP email server exposes mailbox operations as those tools. TempMaily's server exposes six: create_inbox, list_inboxes, delete_inbox, list_messages, get_message, and delete_message. Point your client at the server URL with your API key and the agent can manage real inboxes by itself.
How does my agent read a verification code?
After using its inbox address in a signup, the agent calls list_messages until the verification email appears, then get_message to fetch the full body. The message includes both bodyText and sanitized bodyHtml, so the agent can extract a six-digit code or a confirmation link directly from the content it just read. The same flow works over the REST API for scripted pipelines that aren't agent-driven.
Do I have to use MCP, or is there a plain API?
Both exist and share the same key. The MCP server is the zero-glue path for agent clients — add the config once and the tools just appear. The REST API covers everything else: CI pipelines, test suites, and custom agent frameworks that speak HTTP. They expose the same six operations over the same inboxes, so you can prototype in an MCP client and ship the same flow as REST calls.
Is agent email access free?
API and MCP access are part of Premium, which is one flat plan at $9.90 per month or $90 per year — there is no separate enterprise tier for API access. The free tier is the anonymous browser inbox on the homepage, which is great for trying the product but has no programmatic access. One Premium key authenticates both the REST API and the MCP server.
Which AI clients work with the MCP server?
Any MCP-capable client. Claude Desktop, Claude Code, and Cursor are the common ones — each takes the same small JSON config naming the server URL and your Bearer key. Custom agents built on frameworks with MCP support connect the same way, and agents without MCP support can use the REST API directly instead.