Deploy without leaving your AI agent: MCP deploys from Claude Code & Cursor

The gap between building and shipping

Building an app inside an AI agent feels seamless — you describe what you want, the code appears, you iterate in the same window. Then you go to deploy, and the seam shows: you leave the agent, open a hosting dashboard, and suddenly you're learning about build commands, environment variables, and ports. The flow you were in is gone.

MCP — the Model Context Protocol — closes that gap. livemy.app runs an MCP server that plugs into agents like Cursor and Claude Code. Once it's connected, deploying is something you ask for in the same chat where you built the app: “deploy this,” and a minute later you have a live URL. No tab switch, no new vocabulary.

How it works, in one paragraph

An MCP server exposes tools an agent can call. livemy.app's MCP server exposes deploy tools — push this project, show logs, redeploy, stop or delete it, connect a custom domain, read and write project files. When you tell your agent to deploy, it calls those tools on your behalf, livemy.app builds and hosts the app, and the agent reports back the URL. You're talking to your agent; your agent is talking to livemy.app. (If you want the other direction — hosting your own MCP server — see how to host an MCP server.)

One-time setup

You connect the livemy.app MCP server to your agent once. The shape is the same everywhere; the menu location differs.

• 1. Get your token. In your livemy.app dashboard, create a deploy token. This is what authorizes the agent to deploy to your account.

• 2. Open your agent's MCP settings. In Cursor, that's the MCP servers section of settings; in Claude Code, your MCP configuration.

• 3. Add the livemy.app server. Point it at livemy.app's MCP endpoint and paste your token. Save.

• 4. Confirm the tools loaded. The agent should now list livemy.app's deploy tools as available. You're connected.

That's it — once per machine. From here, deploying is conversational.

The workflow

After setup, you stay in the agent for the whole loop:

• Deploy: “Deploy this app.” The agent pushes the current project to livemy.app and returns the live URL.

• Logs: “Show me the deploy logs” or “why did the build fail?” — the agent pulls logs and can suggest a fix, then redeploy.

• Custom domains: “Connect mydomain.com to this project” — domains and project files are manageable from the chat. (Environment variables and secrets are set in the dashboard.)

• Manage: “Redeploy this,” “stop it,” or “delete the project” — if a deploy misbehaves, you handle it in a sentence.

The point isn't novelty — it's that the debugging loop (deploy → read error → fix → redeploy) collapses into one place. The agent that wrote the code also reads the logs and ships the fix.

Who this is for

If you live in Cursor or Claude Code, this removes the most annoying context switch in your day. If you're less technical and the agent is doing most of the work, it means you never have to understand a dashboard — you just keep asking the agent for what you want, including “make it live.” Either way, the app ends up on a real URL with SSL and a flat $10/month bill, no metering.

FAQ

Which agents can deploy through livemy.app's MCP server?

Any MCP-capable agent. Claude Code and Cursor are the common ones; the connection is by URL and token, so anything that speaks MCP can use the deploy tools.

Is deploying through MCP secure?

Yes — the agent authenticates with a token you generate and can revoke from your dashboard, and all traffic is over HTTPS. The token scopes what the agent can do to your account.

Do I still get a dashboard?

Yes. The MCP connection is an additional way to deploy, not a replacement. Everything you do through the agent shows up in your livemy.app dashboard, and you can manage projects there too — environment variables and secrets live there.

What's the difference between this and hosting an MCP server?

This is deploying your app through an MCP connection — the agent ships your project. Hosting an MCP server means running your own server so agents can call its tools. Different directions; see how to host an MCP server.

Can I deploy an app I built with Lovable or Bolt this way?

If you're working with the exported code in an MCP-capable agent, yes. You can also deploy those directly — see deploy a Lovable app and deploy a Bolt.new app.

Ship from where you build

The best deploy step is the one you don't have to leave your work to take. Connect livemy.app's MCP server once, and “make it live” becomes part of the same conversation that built the app.

→ Get your livemy.app deploy token · Connect it to Claude Code or Cursor · Flat $10/month, no metering.