Keep your computer awake
while your AI agents work
You lock the screen, you go for lunch, and the build you started is still on minute 3 when you come back. AgentsRoom stops that: while an agent works, a dev command runs or a database export is in flight, your computer does not go to sleep.
The screen still turns off. Only system sleep is held, and it is released the moment nothing is running. A cup icon in the status bar tells you when the machine is held and why, and one click keeps it awake on demand, with nothing running, for 1, 4 or 8 hours or until you turn it off.
- Agentrefactoring the auth module
- Dev commandnpm run build
- Databaseexport, 2.1 GB of 3.4 GB
- Manual holdkept awake for 4 h
What the cup in the status bar follows: an agent mid-turn, a build that has not returned its prompt, a database export, or a hold you started yourself. When all four are quiet, the machine is free to sleep again.
Laptops and desktops are set to sleep when nobody touches them. On a recent Mac with factory settings, the system sleeps about one minute after the screen turns off, and locking the screen turns it off right away. Windows and Linux have the same idle timer. For a person reading email, that is the right call. For a machine that is running an AI agent, a build or a database dump, it means every one of those processes freezes in the middle of its work.
We measured it on a real session: three builds started, the user locked the machine and left. The system went to sleep four minutes later. A 13-minute build took 62 minutes of wall-clock time, 45 of them spent asleep, and nothing moved until someone pressed a key. The agent waiting on that build froze with it, and a database export over an SSH tunnel lost its connection.
AgentsRoom now asks the operating system to stay up while it has real work in progress, and to let go as soon as it does not. It is the same request Caffeine, Amphetamine or PowerToys Awake make, except AgentsRoom knows when the work starts and when it stops, so you never have to remember to switch it on, or off.
And when you want the machine up with nothing running, for a long download, a presentation or a remote session, the cup in the status bar does that too, with a timer.
What it does
Automatic while your work runs, manual when you want it, visible all the time.
Holds sleep while work runs
An agent mid-turn, a dev command whose job is still running, a database export, import or copy: any of them keeps the machine up. When the last one finishes, sleep comes back.
An open terminal is not a running one
A dev command tab left open after the build finished does not hold anything. AgentsRoom looks at whether the terminal's job is still running, so npm run dev holds the machine and a finished build does not.
A cup in the status bar
The cup lights up while the machine is held awake. Hover it to see why, click it to see each reason: which agent, which command, which transfer, which manual hold and until when.
Keep awake on demand
1 h, 4 h, 8 h or until you turn it off, even when nothing is running and even if the automatic mode is off. Pick another duration to replace the current one, or turn it off early.
Screen on, only if you ask
By default the screen still turns off and locking stays private. A manual hold can keep the screen on too, which some Windows laptops need, because their standby starts as soon as the screen goes dark.
Agents can ask for it
Through the AgentsRoom MCP, an agent can check whether the machine is held, keep it awake for a set number of minutes when you ask it to, and stop the hold. The cup shows that an agent started it.
How it works
One request to the operating system, made and withdrawn at the right moments.
- 1
AgentsRoom watches its own work
It already knows which agents are working, which dev command terminals have a job running and which database transfers are in progress. That is the whole list of reasons to stay awake, plus a manual hold if you started one.
- 2
It asks the system to skip sleep
While any reason is true, AgentsRoom holds a system power request: a power assertion on macOS, a system-required request on Windows, a desktop inhibitor plus a logind sleep lock on Linux. The screen keeps its own timer.
- 3
It lets go the moment work stops
When the last reason ends, the request is withdrawn and your usual sleep settings apply again. A timed manual hold ends on time, even if the machine was put to sleep some other way in between.
- 4
Nothing can stay stuck on
If the AgentsRoom window crashes or reloads, its report expires after a few minutes and the request is dropped. When AgentsRoom quits, every hold is released, and a manual hold is not restored at the next launch.
When it matters
The moments where a sleeping computer costs you an hour.
Lock the screen and walk away
Start a long agent task, lock your computer, go to lunch. The screen turns off, the agent keeps working, and the result is there when you come back.
Long builds and test suites
A release build, a full test run, a Docker image: the command keeps running with the screen off, and the machine is free to sleep once the job is done.
Database dumps over a tunnel
An export through SSH or AWS SSM can take hours, and the tunnel dies with a sleeping machine. The transfer holds the machine for its whole duration, even after you close the window that started it.
Presentations and remote sessions
Keep the machine and the screen on for an hour while you present, share your screen, or reach this computer from somewhere else, then let it sleep again on its own.
One app less than Caffeine, Amphetamine or PowerToys Awake
Those apps keep the machine awake for as long as you tell them to. They do it well, and AgentsRoom does the same thing with its manual hold: a duration, or until you turn it off, with the screen on or off.
What they cannot know is when your agents, builds and database transfers start and stop. AgentsRoom does, so the automatic hold covers your work without you thinking about it, and lets the machine sleep the minute it is done, instead of keeping it up for hours after the job finished.
If you also keep the machine awake for things that have nothing to do with AgentsRoom, a dedicated app is still the right tool for that. Both can run side by side without conflict.
Your agents can keep the machine awake too
Ask an agent to "keep the computer awake for two hours while the migration runs" and it calls the AgentsRoom MCP. It can read the current state, start a hold for 1 to 1440 minutes or until stopped, keep the screen on if needed, and end the hold. It uses the same hold as the status bar, so you see it in the cup and can stop it with one click. Agents do not need it for their own work: that is already covered automatically.
What it cannot do
- A closed laptop lid still puts the machine to sleep, on macOS, Windows and Linux. No app can prevent that. Keep the lid open with the screen off, or use an external display.
- On Linux, AgentsRoom asks both the desktop session and systemd-logind. A system without systemd, or a sandboxed install that cannot reach it, depends on the desktop session alone.
- On Windows laptops whose standby starts when the screen turns off, only a hold that keeps the screen on keeps them working. The manual hold offers that option and turns it on by default on Windows.
- Everything stops with AgentsRoom: quitting the app releases every hold, and a manual hold is not restored at the next launch.
Frequently asked questions
Why does my computer go to sleep while Claude Code or another AI agent is still working?
Because the operating system only sees that nobody has touched the keyboard. Sleep timers are based on user activity, not on what processes are doing, so an agent, a build or a dump running in a terminal does not count. AgentsRoom tells the system that work is in progress, so sleep waits until it is done.
Will my screen stay on?
No, not by default. Only system sleep is held: the screen turns off on its usual timer and locking your computer works as before. A manual hold can keep the screen on as well if you ask for it.
Can AgentsRoom replace Caffeine, Amphetamine or PowerToys Awake?
For keeping the machine awake, yes. Click the cup in the status bar and pick 1 h, 4 h, 8 h or until you turn it off, with the screen on or off. On top of that, AgentsRoom holds the machine automatically while your agents and commands work, which those apps cannot see.
How do I know if my computer is being kept awake?
Look at the cup icon in the status bar at the bottom of AgentsRoom. It lights up while the machine is held, and hovering or clicking it lists the reasons: an agent working, a dev command running, a database transfer, or a manual hold with its end time.
Does it keep my computer awake forever if I leave a terminal open?
No. An open dev command tab does not hold anything once its job is done. AgentsRoom checks whether the terminal is actually running something, so a dev server that never returns holds the machine, and a build that finished does not.
Does it work with the lid closed?
No. A closed laptop lid puts the machine to sleep on every operating system, and no application can override it. Leave the lid open with the screen off, or connect an external display.
Does it work on Windows and Linux?
Yes. AgentsRoom uses the native request of each system. On Linux it also takes a systemd-logind sleep lock, so it holds on desktops that ignore the usual request. On Windows laptops whose standby starts when the screen turns off, use a manual hold with the screen kept on.
Can I turn the automatic mode off?
Yes, in Settings > Terminal > Keep the machine awake. The manual hold from the status bar keeps working even when the automatic mode is off, and the cup can be hidden from the status bar like any other element.
Can an AI agent keep my computer awake?
Yes, when you ask it to. The AgentsRoom MCP gives agents three tools to read the state, start a hold for a number of minutes or until stopped, and end it. The hold shows up in the status bar as started by an agent, and you can stop it there.
Does it cost anything?
No. Keeping the machine awake, automatically or on demand, is part of AgentsRoom on every plan, including the free one.
You may also like
Process Guard
Your machine slows down after a day of AI coding agents? It is usually not the agents, it is the commands they left behind: one process holding gigabytes, or a dozen holding a little each until nothing is left. AgentsRoom finds them, names the agent that started them, and lets you end them or interrupt that agent in one click.
Scheduled Tasks: cron for AI agents
Schedule recurring AI agents and development workflows without writing a single cron expression. Run a Claude Code or Codex session every morning to review pull requests, scan crash logs every night, check dependencies every Monday, generate a weekly changelog or run a QA pass after every deploy. Pick a frequency (every N minutes, hourly, daily once or twice a day, weekly, monthly, or on your AI plan's quota window), a prompt and an agent role. AgentsRoom launches it for you, catches up missed runs on next launch and keeps a reopenable run history.
Terminal • Detachable • Mobile
AgentsRoom Dev Terminals is a true per-project terminal manager for macOS, Windows and Linux. Launch your backend, frontend, mobile bundler, database and microservices in one click. Detachable window for a second monitor, AI-powered command generation with Claude, real xterm.js terminals, git-native storage, and remote command triggering from the iOS and Android mobile app over 4G/5G. The modern alternative to iTerm2, Terminal.app, Warp, Ghostty, Kitty, Alacritty, WezTerm, Hyper and Tabby.
Database Connections
Save your MySQL, PostgreSQL and MongoDB connections, reach a database in a private subnet through an SSH tunnel or an AWS SSM session, and let your AI agents query it over MCP. Read-only by default, one statement per call, capped result sets, and the password never reaches the agent. Export, import and database-to-database copy included.
MCP : Agent-pilotable : Manual or automated
Four MCP servers let Claude Code, Codex and other CLIs drive AgentsRoom itself: read the backlog, run dev commands, use the prompt library, pilot the browser.
Stop losing builds to sleep
Download AgentsRoom and let your agents finish their work while you are away from the keyboard.
Companion app: monitor your agents on the go
Bring your own: Claude, Codex, Antigravity CLI, or other AI provider.
Push bugs and requests straight to your public backlog.