Cyberun turns the GPUs you already have — owned or rented — into one fleet your team and your AI agents can drive. Run ComfyUI, Nerfstudio, vLLM, Ollama, or your own services, managed, on-prem, or fully air-gapped.
One login, one memory, one fleet — driven from a browser, a terminal, or an AI client.
However your team works, there's a path to your first task — no sales call, no setup project.
Sign in, create a team, connect a GPU, and run a workflow. Most people start here.
Start freeAuthenticate, run workflows, and share a local GPU straight from your terminal or CI.
$ curl -sL https://releases.cyberun.cloud/cli/install.sh | bashPoint Claude Code, Cursor, Windsurf, or Codex CLI at one URL — they list workflows, run tasks, and stream results back.
$ npx skills add cyberun-cloud/skillsYour AI agent lists workflows, submits tasks, streams progress, and calls your team's own services — choosing the right tool itself.
One credential maps to one team. Revoke a misbehaving client in a click, without touching anyone else's access.
Long-running renders stream progress back to your agent as they happen.
One login, one memory, one fleet across every surface. No glue code, no re-auth, no copy-paste between tools.
Sign in once and bring your own single sign-on. The same identity follows you across the dashboard, the CLI, and your AI agents.
Workflows, models, and results live in one place your whole team shares — not scattered across laptops and chat threads.
Every GPU you connect, owned or rented, runs the same workflows and serves the same services — ComfyUI, Nerfstudio, vLLM, Ollama, or your own apps — scheduled as one pool.
Most AI startups rent identity, networking, and scheduling from someone else, and inherit every outage and limit. We built the whole stack ourselves — from bare hardware to the API — so identity, scheduling, and secure connections just work across every site you run. One team owns every line between the GPU and the result.
Read the platform docsEvery box in this diagram exists in code and runs in production today — not a roadmap drawing.
If your work can't live on someone else's GPU, Cyberun is the control plane that makes your own hardware feel like one platform.
Pool a workstation, a rack of servers, and rented cloud GPUs into one fleet — no per-machine setup, no open ports.
Run on-prem or fully air-gapped, with your data and audit trail never leaving your network. No phone-home.
Standardize how your whole team uses AI, and keep every client's work on the GPUs you control.
Give your product teams a paved road to ship AI features, without each one rebuilding identity, scheduling, and storage.
What air-gapped, regulated, and sovereign deployments actually require — built in, not bolted on.
Capability by capability — what's live now, and where it's headed.
| Capability | Cloud | Studio |
|---|---|---|
| Identity & access | ||
| OIDC / OAuth federation | ||
| RBAC + scoped API keys | ||
| Audit log | ||
| Storage & state | ||
| Bring-your-own object storage | ||
| Production asset library | · | |
| Workflow versioning | ||
| Compute | ||
| GPU fleet orchestration | · | |
| Local-first execution | ||
| Air-gap deployment | ||
| External AI provider bridge | ||
| Workflow runtime | ||
| ComfyUI-compatible graphs | ||
| Nerfstudio pipelines | · | |
| Container services | · | |
| Custom node distribution | ||
| Creative production | ||
| Asset-First character lock | · | |
| World / scene composition | · | |
| Multi-layer dubbing (SFX / BGM / TTS) | · | |
| Director-brain critique | · | |
| Delivery & ops | ||
| Signed task webhooks | · | |
| Managed cloud hosting | ||
| On-prem self-host | ||
Create a free team and run your first workflow in minutes — no card, no quotas. Managed, on-prem, and air-gapped plans are available when you need them, with a security review on request.
sales@cyberun.cloud
Get started
sales@cyberun.cloud