DexBox is open source. Spin up computer-use enabled Windows desktop locally.
GitHubGive your agents a real Windows environment. Browser, apps, filesystem, all of it.
Choose your approach
For developers building their own agents. Windows cloud desktops via API. You bring the logic.
Key Functions
For teams that need desktop automation without building the agent themselves. We engineer and run it.
Key Functions
Infrastructure
A Windows desktop in under 2 seconds.
Spin up isolated Windows cloud desktops via API. You bring the agent. Nen handles the desktop.
$ nen desktop create ✓ Desktop ready (1.8s) ID: dsk_8x4k2p9b Endpoint: api.getnen.ai/desktops/dsk_8x4k2p9b Tools: computer, bash, files
from nen import Desktop from anthropic import Anthropic desktop = Desktop("dsk_8x4k2p9b") agent = Anthropic() while not desktop.is_done(): state = desktop.screenshot() action = agent.decide(state) desktop.execute(action)
01.
One API call returns a fresh, isolated Windows desktop in under 2 seconds. No warm pools, no queue, no infrastructure to manage.
02.
Point any computer-use model at the session, whether Claude, GPT, or your own. Screenshots, clicks, keystrokes, and shell commands over a single SDK in Python, TypeScript, or HTTP.
03.
Run thousands of desktops concurrently. Pay only for active session time and tear down when the agent is done. No idle cost, no per-seat licensing.
Fully Managed
Describe the workflow. We engineer it. Your system calls it.
No platform license. No RPA specialists. No six-month implementation. You describe the task, we build and operate it, your system calls one endpoint.
$ curl -X POST api.getnen.ai/runs \ -d @retrieve_chart.json ✓ Run started · run_3kx9p2m4

← POST /api/nen-webhook { "run_id": "run_3kx9p2m4", "doc_url": "api.getnen.ai/d/9f2a" }
01.
No agent installation on the target machine. Works behind firewalls. Works with existing IT. One API call to establish a session.
02.
Define everything in code. No DSL to learn. Computer vision handles the UI. Your logic runs alongside the desktop session to minimize latency and maximize reliability.
03.
Fire automations from any backend, workflow engine, or orchestration layer. Schedule, webhook, or pipeline step. Results come back as structured JSON.
Platform capabilities
Environment
Execution
Platform
Use Cases
Logistics
Computer Use Desktop
Agents handle repetitive data entry in the desktop software your customers already use. No system replacement required.
Healthcare
Managed Workflows
We build and run agents that navigate Windows desktop EHR environments to retrieve documents for appeals and prior authorization workflows.
Financial Services
Computer Use Desktop +
Managed Workflows
Agents complete structured submissions with full traceability. Bring your own agent, or let us build and operate it.