Keep your laptop awake during a long AI agent run.

Tools that run locally — Claude Code, Cursor, GitHub Copilot's agent mode, a local Ollama or LM Studio model — execute directly on your machine's CPU/GPU. The only thing that pauses them is your system actually sleeping (suspending). A locked or dimmed screen doesn't touch them; they keep running in the background exactly the same either way.

Tools that run in the cloud — ChatGPT in a browser tab, Claude.ai, most API-based assistants — do the actual computation on the provider's servers. Your laptop sleeping doesn't pause that computation at all; it just disconnects your view of it. See ourChatGPT-specific pagefor that distinction in more detail.

There are two separate things that can get in the way of a long local run, and they need two separate fixes:

• Screen locking/dimming — annoying (you have to keep unlocking to check on it), but doesn't actually pause the agent. This tool fixes it: no install, works even on a locked-down work laptop.
• Full system sleep — this genuinely pauses everything, agent included, until the machine wakes back up. No browser-based tool can prevent this — the Wake Lock API's spec explicitly dropped the "system" lock type that would have allowed it. Fix this in your OS's own power settings instead: on Windows, Settings → System → Power & sleep → Sleep → Never (while plugged in); on Mac, System Settings → Battery → Options → "Prevent automatic sleeping on power adapter," or run caffeinate -s in a spare terminal tab for the duration of the run.

For a genuinely unattended long run — walk away, come back in a few hours — both matter: the OS sleep setting is what actually guarantees the agent keeps working, and this tool is what means you come back to a usable screen instead of a locked one.