How do I add an MCP server to Claude Code?

Install Clize (npm i -g @clize/clize), run clize login, then clize install --mcp. It registers Clize as a Claude Code MCP server (and a Skill). The manual equivalent is claude mcp add clize -- clize-mcp.

How do I verify the MCP server is connected to Claude Code?

Run claude mcp list. Clize should appear among the registered MCP servers. Start a new Claude Code session and the clize tools — domains, email, deploy, status — become callable.

What is the difference between the Clize Skill and MCP server in Claude Code?

clize install adds a Skill by default — the trigger layer Claude Code reaches for, with the safety rules. Adding --mcp also registers the MCP server, the structured tool layer Claude Code discovers automatically. Most setups use both.

Add Clize MCP server to Claude Code (step by step)

A four-step guide to adding Clize as a Claude Code MCP server — so Claude Code can act outside the sandbox: domains, inboxes, deploys, and support. For the overview, see Claude Code MCP server.

Before you start

You need Node.js (for the npm install) and Claude Code already installed. Clize is the capability layer — it needs no model API key of its own.

1. Install Clize and log in

Install the CLI globally, then authenticate. Login uses Clize's hosted layer, so there are no keys to paste.

$ npm i -g @clize/clize
$ clize login

2. Add Clize as a Claude Code MCP server

One command wires Clize into Claude Code as both a Skill and an MCP server. The --mcp flag is what registers the structured tool layer (without it, Clize installs only as a Skill).

$ clize install --claude --mcp

Prefer to configure it by hand? That's the exact registration the installer runs:

$ claude mcp add clize -- clize-mcp

Only need one product line? A profile registers just that line's tools:

$ claude mcp add inbox -- clize-mcp --profile inbox   # 13 tools instead of 32

Profiles: inbox (13 tools), storefront (15), sites (5), domains (9) — a context budget, not a permission boundary.

3. Verify the connection

Ask Claude Code which MCP servers it knows about. Clize should be listed.

$ claude mcp list

Start a new Claude Code session and the Clize tools — domains, email, deploy, status — are callable. The Skill is the trigger layer; the MCP server is the structured tool layer; the same actions also run as a stable CLI.

4. The first thing to run

With Claude Code connected, the first thing most people run: give the agent an address of its own (clize claim <slug> --email), let it sign up for a service with that address and read the verification code from its own inbox (clize email inbox --wait-for <text>), then answer whoever writes back — the agent drafts, the mail leaves only after you approve (clize email send … --confirm). A support@ desk is one tag on that same inbox — see AI customer support agent. To ship a site instead, see AI agent deployment.

The safety gates

Because these are real actions, 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.

Add Clize MCP server to Claude Code