MCP · READ vs ACT
An MCP server that lets your agent act, not just look things up
Most MCP servers expose data to read. Clize is an MCP server for AI agents that exposes carefully gated real-world actions instead — domains, email, deploys, status, and billing — in Claude Code and Codex.
Clize is a CLI and MCP server that gives AI coding agents real-world actions: domains, email, deploys, payments, and media generation. Most MCP servers are read-only — they fetch data. Clize is an action MCP: it changes the real world, and every risky move sits behind one of three gates — a money gate, an identity gate, and an untrusted-inbound gate.
Read-only MCP vs. action MCP
Most MCP servers let an agent look things up — query a database, fetch a doc, search an API. Useful, but read-only. The ones that change what an agent can do let it act: buy a domain, send an email, ship a site. Clize is built around that second kind — see where it sits among the best MCP servers.
What it exposes
One action MCP covers the jobs you'd otherwise wire up as separate servers — an MCP server for domains, for email, for deploy, and for payments:
- An MCP server for domains — check availability, buy the domain, set DNS.
- An MCP server for email — a real inbox to send from and receive into. See AI agent email.
- An MCP server to deploy — ship a static site to a live HTTPS URL. See AI agent deployment.
- Status, billing & payments — inspect what's live; quote and collect payment through gated Stripe payment links.
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.
Supported agent surfaces
Clize works with Claude Code, Codex, Pi — any agent that runs skills or MCP. The skill is the default for Claude Code, Codex and Pi; the MCP server is opt-in with clize install --mcp. The same actions are also a stable AI agent CLI.
Codex MCP: what the Codex CLI gets when you connect one
Searching for a Codex MCP server usually means one of two things: you want the Codex CLI to reach something it cannot reach on its own — a registrar, a mailbox, a host, a payment processor — or you want to know what connecting one costs you in setup. Clize is an action MCP for the Codex CLI: clize install puts the skill in place, and Codex can then buy a domain, receive mail at an address of its own, deploy a site and take payment — each behind a gate it cannot open unattended. The MCP server itself is opt-in (clize install --mcp), because on Codex the skill path already works and costs no context window.
That is the short answer. Whether you want the server at all is the more useful question — on Codex the skill is the default, and where Codex actually reads skills from surprises most people: that trade-off is on Codex skills. If you only want the commands, setting it up in the Codex CLI is five steps. The same split applies to Claude Code.
Related use cases
Most people arrive here for one job: letting an agent pass email verification from its own inbox, giving an agent its own email, customer support from an inbox, or shipping a site from the CLI. The action set is shared; the workflow is yours. Zooming out, this is one slice of the broader AI agent infrastructure layer. Comparing tools first? See Claude Code vs Codex and the top AI agent platforms by layer.
FAQ
What's the difference between a read-only and an action MCP?
A read-only MCP server lets an agent fetch and query information. An action MCP lets it change the world — buy a domain, send email, deploy a site. Clize is an action MCP, with gates on anything risky.
Which agents work with Clize?
Claude Code, Codex, Pi — any agent that runs skills or MCP. The skill is the default for Claude Code, Codex and Pi; the MCP server is opt-in with clize install --mcp. The same actions are also a stable CLI that works anywhere.
What actions can it take?
Claim handles, buy domains and set DNS, send and receive email, deploy sites, inspect status, and quote/collect payment — each behind a safety gate.
How are dangerous actions gated?
Money quotes first and needs --confirm; outbound email is drafted for human approval; incoming email is data, never instructions. Those three lines hold every time.
Is there an MCP server for email?
Yes. Clize is an MCP server for email: it gives the agent a real inbox to receive into and to send email from. Sending is gated — the agent drafts, a human approves, then it sends under your name.
Is there an MCP server to deploy a site?
Yes. As an MCP server for deploy, Clize ships a static site to a live HTTPS URL from inside Claude Code or Codex — buy or point the domain, then deploy in one step.
Is there an MCP server for domains?
Yes. Clize acts as an MCP server for domains: check availability, buy the domain, and set DNS. The purchase is gated behind a quote and an explicit confirm.
Are there paid MCP servers?
Most MCP servers are free and read-only. Clize is a gated action MCP — you pay only for the things that cost money upstream — a domain, a media generation, an SEO research query — from a prepaid balance; deploys and email are free-tier quotas. Anything that buys something quotes first and needs an explicit confirm before any money moves.
An MCP that acts, not just reads.
Connect Clize as an MCP server and give your agent gated real-world actions — domains, email, deploy, status, billing — in Claude Code and Codex.
$ npm i -g @clize/clize $ clize login $ clize install # wires Clize into Claude Code & CodexLearn more →