Le Bureau vs Bytebot: managed or self-hosted desktop?

If you want to give an AI agent a real computer instead of a throwaway sandbox, two names come up: Bytebot and Le Bureau. Both give an agent a full desktop with a browser, a terminal, and a file system that persists between runs. The real difference is who runs the machine.

The short version

• Bytebot is open-source and self-hosted. You clone the repo, add your AI provider key, and run it with Docker on your own hardware. You own the desktop, and you own the operations.
• Le Bureau is the managed cloud version. Each agent gets its own isolated virtual machine, hosted in Europe, with nothing for you to install or maintain. You get the same real desktop without running any infrastructure.

Both are a better fit for real agent work than an ephemeral sandbox, because the environment survives between sessions. The question is whether you want to operate that environment yourself.

What Bytebot is good at

Bytebot is a solid open-source project. Because it runs on your own machine, you get full control over the host and the network, no per-seat cost beyond your own compute, and the ability to read and modify the code. That makes it a good choice for a developer who wants to tinker, run a single desktop locally, and keep everything on hardware they control. If self-hosting is the goal, Bytebot is built for it.

The trade-off is that you are now running the infrastructure. You maintain the host, the storage, the updates, and the isolation between desktops. A single local desktop is easy. A fleet of them, one per agent, with real isolation and uptime, is a different job.

What Le Bureau does instead

Le Bureau treats the desktop as a managed service. You do not install anything. You create an agent, and it gets its own dedicated virtual machine with a persistent disk, isolated from every other agent. It has a real browser, terminal, editor, and file system, streamed to your browser. Mission Control gives you one place to manage every agent, task, and desktop. Hosting is in Europe, with data resident in the EU and aligned with GDPR.

Because each agent runs in its own isolated machine, a mistake or a compromise in one agent cannot reach another. You get that isolation without designing it yourself. The trade-off is the mirror image of Bytebot: you do not own the host. In exchange, there is nothing to operate.

Side by side

How to choose

Pick Bytebot if you want to self-host, you are comfortable running Docker and maintaining the environment, and one desktop on your own hardware covers your needs.

Pick Le Bureau if you want the same real desktop for your agents but would rather not run any of it. You want per-agent isolation, a team view, and European hosting out of the box, and you would rather spend your time on the agent's work than on keeping the machines healthy.

Both approaches beat handing your agent a sandbox that forgets everything after each run. The choice is about ownership, not capability.

Related reading

• Le Bureau vs Claude Cowork: the managed-cloud comparison against Anthropic's desktop agent.
• E2B vs Browserbase vs Scrapybara vs Le Bureau: sandboxes vs a full desktop.
• Cloud desktops for AI agents: the full guide.

Try it

You can create your first agent on Le Bureau for free, and it gets its own cloud desktop in about two minutes. Bring your own AI provider key, or use a managed model, and see how your agent works when it keeps its environment between sessions.