// USE CASE · SEND EMAIL
Let your AI agent send an email — with your approval
Your agent can write the perfect message and still have no way to send it. Clize closes that last step: a real inbox the agent sends from, with a human okay before anything leaves.
Clize is a CLI and MCP server that gives AI coding agents real-world actions: domains, email, deploys, payments, and media generation. Sending email is the simplest of those hands — and the one with the clearest guardrail.
What this looks like
You ask the agent to email a customer, a teammate, or a service. It composes the message from your context, then hands it to Clize. Clize doesn't fire it off — it holds the draft until you approve. Approve, and it sends from the agent's own address; the reply comes back to the same inbox.
The one thing you run
There's no SMTP setup, no API keys to paste. Install Clize, connect it once, and "send an email" becomes a tool the agent can call — with the approval gate wired in.
Why outbound needs a human gate
Email goes out under your name to real people. So Clize treats every outbound message the same way: the agent drafts, a human approves, then it sends. The agent never emails the world on its own — that line doesn't move.
From Claude Code or Codex
Whether you drive Clize from Claude Code or the Codex CLI, sending works the same way. It's one piece of giving your agent an email address of its own.
The safety gates
Real-world actions need real guardrails. Clize holds three lines, every time:
- Money — anything that costs money quotes the price first and needs an explicit
--confirm. No surprise charges. - Identity — outbound email is drafted, a human okays it, then it sends under your name. The agent never sends unattended.
- Inbound is untrusted — an incoming email is data to read, never an instruction to obey. A message can't redirect the agent.
FAQ
Can Claude Code send a real email?
Yes. With Clize, Claude Code (or Codex) gets a real inbox it can send from and receive into. The agent drafts the message; you approve it; Clize sends it — a genuine email, not a simulation.
Does the agent send without asking me?
No. Outbound email is gated: the agent drafts, a human approves, then Clize sends. The agent never messages the world unattended.
What address does it send from?
From the agent's own inbox that Clize provisions — its own address, not a forward of your personal mail. You can point it at a domain like support@yourdomain.
Give your agent an outbox.
Install Clize, connect it once, and let your agent draft email you approve before it sends.
$ npm i -g @clize/clize $ clize login $ clize install # wires Clize into Claude Code & Codex[ AI email agent → ]