Glasswarp FAQ — eyes and hands on real Windows
Quick answers about the API, pricing, and your Windows PCs.
What does Glasswarp provide vs what I build?
Glasswarp is eyes and hands: native-resolution vision, native input, and a safe session (consent, Live View, kill switch, audit). You bring the brain — the model, prompts, and task logic.
How do I get started with an agent?
Create an API key, install the Windows host, enable API access on a rig,
then wire the SDK (pip install glasswarp) into your agent loop.
Templates live in the docs. You keep the model and task logic.
Do I need to rent cloud GPUs or Windows desktops?
No. Install the host on Windows machines you already own. You pay Glasswarp for session minutes — not for someone else’s hardware.
Do I bring my own model?
Yes — any model or agent framework. No lock-in. Glasswarp doesn’t include or require a specific LLM; it gives your stack vision and input on a real Windows desktop.
How is this different from VNC or computer-use sandboxes?
Sandboxes usually approximate a desktop. Glasswarp captures a real Windows machine from the GPU pipeline at native resolution with native Win32 input — your apps, files, and GPU.
Is API access safe for the machine owner?
API access is off until the owner enables it per rig. An on-screen “API session active” indicator shows during sessions, owners can kill any session from the console, and every session is audited.
How does pricing work?
Free ($0): 500 minutes/month, 1 concurrent session. Builder: $20/mo or $200/yr — 2,000 minutes, 3 sessions. Growth: $99/mo or $990/yr — 12,000 minutes, 10 sessions. Unlimited paired rigs on every plan. See pricing.