Launch your 24/7 agent in a few hours on a dedicated Mac.
A Mac mini in our rack that never sleeps. Claude Code and Codex are installed and waiting for your sign-in. You open the desktop from a browser tab and hand it work on your code from your phone. It costs $149/mo and you can cancel anytime.
Sign up and pick your agent
Choose Claude Code or Codex and check out. Takes two minutes.
We set up your Mac
Within 1 to 4 hours you get an email with a link to open your Mac in the browser.
Sign in and hand it work
A Terminal is already open with your agent running. Sign in with your own account, then give it a task on your codebase from your phone or the desktop.
Everything an agent needs to run unattended.
A Mac that stays up
Sleep is off, the session stays logged in, and the machine is on wired ethernet with battery backup. Your agent keeps working while your laptop is closed.
Claude Code and Codex, ready
Both are installed. You pick which one runs first. Sign in with your own Claude or ChatGPT account. You pay your usual plan and nothing per token to us.
A real desktop in your browser
Click Open my Mac in your portal and the desktop appears in a tab. Approve a login, paste a code, or watch the agent work. Nothing to install.
Control from your phone
Once signed in, send tasks from the Claude app or the ChatGPT app. Come back to the desktop only when the agent gets stuck on something a human has to do.
Install anything else
It is your Mac. Add Chrome extensions, Homebrew packages, or another coding agent. We set up Claude Code and Codex. The rest is yours to run, on your software.
One operator, real email
A person who knows the machine answers support@macrun.dev, usually within the hour. No ticket queue.
A real M4 Mac mini, all yours.
| Hardware | Apple M4, 10-core, 16GB RAM, 256GB SSD |
| Uptime | Sleep disabled, auto-login, wired ethernet, battery backup |
| Installed | Claude Code, Codex, Google Chrome, tmux, Homebrew, Xcode |
| Access | Browser desktop from your portal. SSH and Screen Sharing over Tailscale if you want them. |
| Your account | A user named runner with admin rights. Install what you like. |
| Price | $149/mo, monthly, cancel anytime. The rate you sign up at is the rate you keep. |
What we do not do.
- −We do not sell API credits. Your Claude or ChatGPT plan covers usage.
- −We do not read your sessions. Sign-in happens on your Mac, in your account.
- −We do not babysit the agent. When it needs a human, that human is you, from your phone or the browser desktop.
- −We do not run other agents for you. Install another coding agent yourself if you want one. It will work, and we will not support it.
- −We do not host general automation. Apple licenses a leased Mac for software development, so agents here work on your code, not your inbox or your CRM.
Need the Mac as a GitHub Actions runner instead? That is the CI runner plan at $119/mo. Same hardware, different job.
Questions people ask before renting
What is an agent Mac?
+
A dedicated M4 Mac mini in our rack that stays on 24 hours a day. Claude Code and Codex are installed and a session is kept alive, so a coding agent can work on your software while you sleep. You open its desktop from a browser tab and control it from your phone. It costs $149 a month, cancel anytime.
Do I need an API key?
+
No. You sign in to Claude Code with your Claude account, or to Codex with your ChatGPT account. Your existing subscription covers the usage. We never see your keys or your conversations.
Which agents are installed?
+
Claude Code and Codex. Both are installed and updated. At signup you pick which one is running when you first open the Mac. You can switch or run both later.
Can I install other tools or agents?
+
Yes. It is your Mac with an admin-capable user. Install Homebrew packages, Chrome extensions, other coding agents, or anything else your project needs. We support Claude Code and Codex directly. Other software is yours to run, and it has to stay on software work like everything else on the Mac.
How do I control the agent from my phone?
+
Once you are signed in on the Mac, the Claude app and the ChatGPT app can send tasks to that session. When the agent needs a human, for example to approve a login or read a two-factor code, open the desktop in your browser and do it there.
How is this different from the CI runner plan?
+
Same hardware, different job. The runner plan turns the Mac into a GitHub Actions runner for $119 a month. The agent plan keeps a logged-in desktop alive for AI agents, adds the browser desktop, and costs $149 a month. You can lease one of each.
Can the agent do things other than software work?
+
No. Apple licenses a leased Mac for software development, building, testing, and the tools that support them. A coding agent working on your own software fits that. An agent that manages your inbox, your CRM, or your social accounts does not, and we will suspend a unit used that way. Every hosted Mac provider works under the same license.
What happens to my agent's files when I cancel?
+
Access ends at the end of the paid period, then the machine is wiped. Copy anything you want to keep before then. The volume is securely erased and the Mac is restored to a clean image before anyone else uses it.
How to run Claude Code 24/7 on a Mac and control it from your phone
What it takes to keep Claude Code running all day on a Mac: a machine that never sleeps, a logged-in session, sign-in with the Claude subscription you already have, and phone control through Remote Control.
Always-on Mac for AI agents: what you actually need to set up
The checklist for a Mac that runs AI agents unattended: power settings, automatic login, a supervised session, screen access for when it gets stuck, wired network and battery backup. Learned on real machines.
Running OpenAI Codex 24/7 on a dedicated Mac
How to keep OpenAI's Codex CLI running all day on a Mac: sign in with your ChatGPT account, keep a login session alive, and give it a real desktop and browser to work in.
Renting a Mac for AI agents vs buying a Mac mini: the actual math
A straight comparison for someone who wants an always-on Mac for AI agents: what a Mac mini costs to buy and run right now, what renting one ready-made costs, and which is right depending on how long you need it and who does the setup.
What an AI agent can and can't do on a hosted Mac
The honest capabilities and limits of running an AI agent on a Mac in someone else's rack: what works, what is unreliable, and what is off the table, so you know before you rent.
Claude Code stops when your Mac sleeps. How to keep it running.
Why a long Claude Code or Codex task dies when your Mac sleeps or the lid closes, the pmset and caffeinate settings that stop it, and the point where a laptop is the wrong machine.
Keep a Claude Code session alive with tmux and pick it up anywhere
The tmux setup for a coding agent that has to outlive your terminal window, your SSH connection, and your commute. Start, detach, reattach from a phone, and recreate the session after a reboot.
Running Claude Code unattended: permissions, guardrails, and what never to allow
How to let Claude Code or Codex work with nobody watching: permission modes, allow and deny rules, hooks, git guardrails, scoped credentials, and the short list of things an unattended agent should never be able to do.
How to let a coding agent work overnight without waking up to a mess
A repeatable way to hand a backlog to Claude Code or Codex at night and review clean pull requests in the morning: how to write the tasks, how to isolate each one, what the agent may do alone, and the ten-minute morning review.
Where to run a 24/7 coding agent: laptop, Linux VPS, EC2 Mac, or a hosted Mac mini
An honest comparison of the places a Claude Code or Codex agent can live all day: your laptop, a Linux VPS, an EC2 Mac, a Mac mini you buy, or a hosted Mac mini. Costs with sources, what each cannot do, and a decision rule.
Why an iOS coding agent needs a real Mac, and the build-run-screenshot loop that makes it useful
What an AI agent working on an iOS or macOS app can only do on macOS: build with Xcode, run the simulator, take screenshots of the result, and sign a build. The exact commands for the loop, and what a Linux box can and cannot do.
Your agent, working tonight.
Sign up now and your Mac is ready within 1 to 4 hours. Sign in once, then run it from your phone.
Launch an agent Mac for $149/mo