// CODEX · MCP
Give the Codex CLI an MCP server for real-world actions
The Codex CLI writes and runs code. Clize is the MCP server that lets it act outside the repo — buy domains, send and receive email, ship sites, and check status, all behind safety gates.
What Codex does alone vs. what Clize adds
Codex is strong inside the sandbox: it reads your repo, writes code, and runs commands. What it can't do on its own is reach the world that code ships to — there's no domain to buy, no inbox to send from, no deploy target it owns. Clize is the Codex MCP server that adds exactly that layer.
Set up the Codex CLI MCP server
Install the CLI and register Clize as a Codex CLI MCP server; Codex picks up the tools automatically. The same capabilities are also available as plain CLI commands, so a step can fall back to a stable command when you want determinism. Once connected, codex mcp list shows Clize among your Codex MCP servers. For a step-by-step walk-through, see how to set up Clize MCP in the Codex CLI.
What your agent can do
- Domains — check availability, buy a handle, point DNS.
- Email — a real inbox the agent can send from and receive into.
- Deploy — ship a static site to a live URL over HTTPS.
- Status & context — inspect what's claimed, deployed, and pending.
- Billing & pay links — quote costs and collect payment, gated.
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.
A typical run
You ask Codex to launch a small site. It generates the pages, Clize buys the domain and deploys it, then tags a support@ inbox pointed at your docs. From then on, Codex can read incoming mail and draft replies — you approve before anything sends.
FAQ
How do I connect an MCP server to the Codex CLI?
Install the Clize CLI, run clize login, then clize install to register Clize as a Codex CLI MCP server. Codex picks up domains, email, deploy, and status as callable tools.
Does the Codex CLI support MCP servers?
Yes. The Codex CLI can connect to MCP servers, and Clize registers as one — exposing real-world actions (domains, email, deploy, status) as tools Codex can call.
What can Clize do that Codex can't alone?
Codex stays inside the repo. Clize gives it hands outside it: buying a domain, sending and receiving email, deploying a site, and checking real-world status — with money and identity gates.
Is it safe to let an agent spend money or send email?
Spending quotes first and requires an explicit --confirm. Outbound email is drafted for a human to approve before it sends. Incoming email is treated as data, never as instructions.
Do I need model API keys?
No. Clize is the capability layer — it does not need an LLM key. Your agent (Codex) is the brain; Clize is the hands.
How do I verify the MCP server with codex mcp list?
Run codex mcp list in your terminal — Clize shows up in the list of MCP servers connected to the Codex CLI, with its domains, email, deploy, and status tools ready to call.
Connect Codex to the real world.
Install Clize, log in, and connect it as an MCP server — Codex keeps writing code and now also ships it: domains, email, deploys, support.
$ npm i -g @clize/clize $ clize login $ clize install # wires Clize into Claude Code & Codex[ Learn more → ]