A sandbox per agent,
with an undo button
Let an agent install, run and break things on a machine of its own. Nothing reaches yours, and a rewind puts everything back.
No newsletter. One email when your access is ready.
What it gets you
A virtual machine per agent
Not a shared container: each agent has its own machine, its own disk and its own network.
Full access inside
The agent can install anything it needs, as an administrator of its own machine.
Rewind
A checkpoint every 15 minutes. One click restores the whole machine in 0.14 s.
Encrypted at rest
Disks and checkpoints are encrypted on our servers in Europe.
Keys stay out of the database
Model keys go to the agent's memory only. They are never stored in our database.
Limits that protect you
Processor, network and email limits keep a runaway agent from hurting anyone.
Let it break its computer,
not yours.
Give the agent full control of its own machine. If an experiment goes wrong, restore a checkpoint and try again.
Questions
and answers.
Keep
reading.
- Running AI agents on Waken vs on your laptopAgents on your laptop stop when the lid closes and can touch all your files. Here is how a Waken agent compares, with its own machine in the cloud.
- AI agent with its own browser and desktopGive an AI agent a Linux desktop with a browser it can see and use. Watch it live, take over for logins, record the screen. Ready in under a second.
- Run Claude Code in the cloud, 24/7Host Claude Code on its own cloud machine. It keeps its files and memory, sleeps when idle, wakes in 0.08 s, and answers from Telegram, Slack or your terminal.
- Run the Codex CLI in the cloudHost OpenAI's Codex CLI on its own cloud machine. It keeps its repository and history, runs on your OpenAI or OpenRouter key, and sleeps when idle.