runlease

Guide ยท MCP

An MCP server that gives your agent a cloud machine

runlease mcp is a Model Context Protocol server. An agent that speaks MCP gets typed tools to launch a machine in your cloud account, run a command on it, copy a file, extend the lease and delete it, without composing shell commands or parsing their output.

Register it

The server is the same binary as the CLI, so there is nothing more to install. It uses the sign-ins the CLI already has. Get one launch working by hand first:

$ curl -fsSL https://runlease.online/install.sh | sh
$ runlease login
$ runlease up -p gcp -n box -d && runlease down box --yes

Claude Code

$ claude mcp add runlease -- runlease mcp

Add --scope user to have it in every project.

Any other MCP client

{
  "mcpServers": {
    "runlease": {
      "command": "runlease",
      "args": ["mcp"]
    }
  }
}

Claude Desktop, Cursor, VS Code and others read this shape.

The client starts the server itself. You never run runlease mcp by hand.

The tools

ToolWhat it does
runlease_upLaunch a spot machine: cloud, optional name, size, lease and region. Returns at once with the machine's details.
runlease_execRun a command on a machine. Returns the exit code, stdout and stderr.
runlease_push, runlease_pullCopy one file to a machine, or from it.
runlease_snoozeSet a machine's self-deletion to a duration from now.
runlease_downDelete a machine now.
runlease_list, runlease_statusThe machines this computer tracks, who is signed in, and how many machine slots are in use. Read-only.
runlease_doctorCheck sign-ins, project, network and relay, and report the fix for each problem. Creates nothing.

Why typed tools

Questions

Where do the machines run?

In your own AWS or Google Cloud account, as spot VMs. Nothing is hosted for you.

Does the server send anything to Runlease?

The CLI has no telemetry. It talks to your cloud and to the relay you chose, and the relay cannot read your sessions.

Is there more for Claude Code than the tools?

Yes. The plugin adds skills that tell Claude when a separate machine is the right call, and a live view of what is running.

Running agents for a whole team? Runlease gives one agent one machine. Marshal, from the same team, runs AI coding agents in isolated pods inside your own VPC, with credential isolation, audited egress and live tool policies.

More guides