Which advantages over its rivals hold up, and why are they hard to copy?
Boxes.dev's advantage is architectural: each agent gets a whole isolated Linux machine with its own filesystem, port forwarding to localhost, and the developer's setup preinstalled, so every meaningful body of work runs on its own cloud machine with no resource constraints or port juggling, and parallel agents never contend for one laptop's ports or resources. It runs native Claude Code and Codex on fresh cloud copies of the developer's existing dev environment. A rival that shares one local machine cannot match that isolation without giving every agent its own machine.
Agents keep running in the cloud after the developer closes the laptop. Idle boxes carry no charge, and metered use is billed by the second.
How does it make money per customer, and what is still unknown?
Boxes.dev meters revenue by the awake box-hour, so the amount it earns per customer grows with how long agents keep running. It charges only for compute: time beyond the included box-hours costs $0.40 per box-hour on the Starter, Pro, and Max plans before prepaid discounts, while sleeping boxes are free and usage is billed by the second.
A box-hour means one hour awake on a standard box with 4 vCPU and 8 GiB of memory, and larger boxes consume more box-hours for the same awake time. Each box also includes port forwarding to localhost and a preinstalled developer setup.
What does the product do for its buyer, and how is it delivered?
Boxes.dev gives each coding agent its own cloud Linux machine running the developer's preinstalled setup with port forwarding, allowing a buyer to run native Claude Code or Codex on isolated copies of that environment and access them from any client surface. Boxes.dev calls itself the first cloud-only agentic dev environment.
Each new devbox boots from a saved snapshot in seconds, and one API call launches an agent on a fresh devbox. Sleeping boxes are free, and usage is billed by the second.
