Roost is a private AI server for your organization, shipped to you preconfigured. Plug in power and network, and start working: no drivers, no model downloads, no command line.
What you get
- Powerful local models: prompts, code and files stay on hardware you own
- Development and agentic tasks run on Roost
- A dashboard: a powerful chat interface prepared for an agentic future
- An isolated container for every user: full development power on a shared machine
- An OpenAI-compatible API for the tools your team already uses
- Up to 500 users, with up to 8 active inference sessions
- A perpetual license, the software as you receive it, forever.
- Human support and updates while your license is active.
Hardware
128 GB: AMD Ryzen AI Max+ 395, 16 cores / 32 threads, Radeon 8060S, 128 GB LPDDR5X.
192 GB: AMD Ryzen AI Max+ PRO 495, 16 cores / 32 threads, Radeon 8065S, 192 GB LPDDR5X, with more memory and room for larger models.
Both configurations include 1 TB NVMe storage. 4.5 L, 3.1 kg. Power draw from 12 W to 180 W. 5 Gbit Ethernet and Wi-Fi 7.
Licensing
We believe in old-fashioned software: you pay for a license, and you get its features for life. Your license covers every version released while it is active. Renew yearly for support and the latest version, including new features and models. If it lapses, Roost keeps running the last version it received. You can renew later to catch up.
The first license year is included. Renewal is $500 per year, covering up to 500 users.
While your license is active and Roost is online, updates apply automatically. Offline sites can request update media on a flash drive. Need more than 500 users? Contact us.
Usually ships within 2-4 weeks. Subject to hardware availability