Batch processing
Video transcoding, image generation, ETL pipelines, any workload that fits a container.
Docker compute
Containers on idle hardware. Cheaper than spot.
Submit any OCI image. We dispatch it to a Linux or Mac provider with free capacity, kernel-isolated via gVisor or Kata. $0.018 / vCPU-hour beats AWS spot, no lock-in. Bring your image, your inputs, your outputs. We bring the iron.
What you can run
ML inference
CPU-bound inference for smaller models. Use the GPU tier for larger models.
Web scraping helpers
Browserless / headless Chrome containers running alongside our proxy network.
Periodic jobs
Cron-equivalent scheduling. Pay only for execution time. No idle cluster cost.
Custom CI runners
Self-hosted GitHub Actions / GitLab runner images. We provide the capacity.
Build farms
Yocto, Bazel, distributed compilers — pin to ARM64 or x86_64, scale to dozens of providers.
Isolation, in detail
- Linux providers: gVisor (Google’s userspace kernel) or Kata Containers (lightweight VM per container). Provider’s host kernel is shielded; container escapes are blocked at the gVisor sentry.
- Mac providers: Each container runs inside Docker Desktop’s lightweight VM. The provider sees CPU + RAM usage but not container internals.
- Resource caps: CPU, RAM, GPU, bandwidth, runtime are all capped per workload. A provider’s machine is never throttled below the floor they configured in their dashboard.
- Network policy: By default, your container can reach the public internet via the provider’s IP and your job’s S3 bucket. Inbound is denied. Outbound to provider’s LAN is denied.
Submit a container
Bring an image. Bring a CLI. Pay per second after the first minute. Logs and exit code stream to your dashboard in real time.
See pricing