A local Paperclip install is the fastest way to see what it does, and the fastest way to discover that a sleeping laptop does not miss one chat, it misses the entire company's work cycle. Heartbeats are how Paperclip schedules everything, and a heartbeat that never fires is an org chart of agents sitting idle until you open the lid.
How do I keep Paperclip running when my laptop is off?
Move it to a machine that stays awake, and remember there are two layers to move, not one. The Paperclip server holds the scheduler, so routines and heartbeats only fire while it is up. The agents it delegates to need to be reachable and running too, whether that is a sandbox provider it provisions or a runtime living somewhere else. Put the control plane on a server or in Docker, and put the agents somewhere that does not sleep either.
The problem
Laptop sleeps, heartbeats stop
What stalls
Routines, goals, delegation
Layers to move
The control plane and the agents
The fix
Hosts that never sleep
What stops when the laptop sleeps
Layer
On your laptop
On a host that stays awake
Paperclip server
Suspends, so no heartbeat fires.
Scheduler keeps running, routines fire on time.
Routines and goals
Silently skipped until you open the lid.
Advance overnight as configured.
Agents doing the work
Unreachable, so delegated work fails.
Reachable and running, on a sandbox provider or a managed runtime.
What you maintain
One device, and its uptime.
A server and a database, or a platform that runs the agents for you.
Heartbeats are the entire scheduling model
Paperclip does not sit and wait for you to type. Agents wake on heartbeats, routines run on a cadence, and delegation flows between agents on that same clock. All of it lives in the server process. When the machine hosting that process suspends, nothing queues up and catches later; the cycle simply does not happen. That is a different failure from a chat app going offline, because the work you were counting on overnight was the whole point.
Why a laptop cannot be the host
The things that make a laptop a good personal computer are exactly what break a scheduler.
Sleep and lid-close suspend the server process and its heartbeats.
Network changes cut the control plane off from the agents it delegates to.
Reboots for updates end the process with no warning and no restart.
Disabling sleep only ties your uptime to one fragile device you also need to carry.
Move the control plane first
Paperclip's own deployment documentation covers the honest options: run it in Docker, or deploy it to a server such as a VPS or Fly, with a real database behind it, storage configured, and secrets held properly. If you want the dashboard private rather than exposed, their docs cover reaching it over a Tailscale network instead of publishing it. This is the part you can do today, and it is a normal server deployment with normal server responsibilities.
Then move the agents, because that is a second machine
This is where most people stop too early. Getting the control plane onto a server makes the scheduler fire, but a heartbeat that wakes an agent whose runtime lives on your laptop has not solved anything. Paperclip can provision execution environments through a sandbox provider, or delegate to a runtime running elsewhere through an adapter. Either way the agents need a home that stays online, separate from the one the control plane sits on.
If the reason you are reading this is that you do not want to operate servers, the agent layer is the one you can hand off today. Qoren deploys OpenClaw, Hermes, and Codex agents into dedicated cloud environments that stay online by default, put them on a schedule, isolate their secrets, monitor their health, and cap their spend, all from a dashboard rather than SSH. Qoren does not host the Paperclip control plane; that part stays yours. But it removes the laptop from the half of the stack that does the actual work.
Because the scheduler lives in the Paperclip server process, and closing the lid suspends it. Heartbeats and routines only fire while that process is running, so nothing wakes and nothing catches up later.
Run OpenClaw or Hermes without managing infrastructure.
Deploy a managed agent environment, configure the runtime, and keep the agent online without Docker, VPS setup, or server maintenance.