Your GPU computers and private services.Ready for your team, wherever they work.

Turn on-premises or cloud GPU computers into secure multi-user servers. Pool them into one resilient cluster so your team can use all available capacity—no ongoing administration required.

• Define policy once; MTGPU registers hosts, creates isolated user accounts, and installs SSH keys.
• Run rootless container workspaces and schedule native or containerized jobs by CPU, RAM, GPU, and VRAM requirements.
• Give users one place to discover authorized hosts, monitor resources, open SSH terminals, manage workspaces, and submit jobs.

Publish AI backends such as Ollama, LM Studio, and vLLM—or private web apps, SSH, remote desktop, Nextcloud, Gitea, and other TCP services—to people you trust.

• Authorize access service by service instead of exposing an entire computer or network.
• Let users see services from multiple administrators in one Portal, with ready-to-run connection commands.
• Keep every service on its existing private network; no application migration is required.

DirectLink handles registration, discovery, authorization, and signaling-assisted peer-to-peer connectivity. It gives invited users secure access without turning your service into a public endpoint.

• No public IP address, domain, VPN rollout, or router port forwarding.
• A direct, end-to-end encrypted data path whenever the network permits—without vendor infrastructure carrying application traffic.
• A lightweight self-hosted relay option for the rare networks that cannot establish a direct path.