Cheapest GPU across your accounts
The cheapest matching spot GPU across every region of every account — often 2× cheaper elsewhere, up to 90% off on-demand.
anycloud scans every region of every account you connect for the cheapest GPU — pool your credits, keep your data.
Your accounts. Your credits. Your data.
curl -fsSL https://get.anycloud.sh | shanycloud gpus --type h100 --spotLower your GPU bill
Same job, cheapest spot GPU across every account you connect — auto-recovering when preempted, so long runs still finish.
anycloud submit --gpu-type h100 --spotOne submit, every region of every account — the lowest price wins. Illustrative spot prices.
The cheapest matching spot GPU across every region of every account — often 2× cheaper elsewhere, up to 90% off on-demand.
On preemption, your job resumes from the last checkpoint on a new VM — most ML frameworks already write one.
Connect cloud accounts with unused balances, then submit work through one flow.
Agent-first
Point Claude Code, Codex, or Cursor at your own cloud — hundreds of experiments a day, every session auto-tagged, caps optional. See Agents.
$ anycloud budget set 4000 --per day --agent-session Budget day (resets in 19h) per agent-session cap $4,000 each train-llama-sft $3,848 / $4,000 $152 left lr-sweep $3,136 / $4,000 $864 left eval-harness $368 / $4,000 $3,632 left
In production
“We have been using anycloud at @Angstrom_ai to make the gpus go brrrrrr!”
Run anywhere
anycloud prepares your input bucket as files under /mnt/input and uploads files written under /mnt/output back to your output bucket.
import anycloud from pathlib import Path from anycloud.types import CloudConfig @anycloud.function( image="ghcr.io/acme/train:latest", gpu="h100:8", cloud_config=CloudConfig( credentials="my-aws", spot=True, input_bucket="training-data", output_bucket="results", ), ) def train(learning_rate: float): data = Path("/mnt/input/data").read_text() # your training loop reads local files Path("/mnt/output/result").write_text( "loss=0.12\n", ) job = train.submit(0.001) job.wait()
anycloud submit ghcr.io/acme/train:latest \ --gpu-type h100 --gpus all --spot \ --credentials my-aws \ --input-bucket training-data \ --output-bucket results \ -- python train.py \ --lr 0.001 \ --data /mnt/input/data \ --out /mnt/output/result
How it works
anycloud handles the VM lifecycle around your container: start compute, run the image, recover from interruption, move outputs, and tear everything down.
Use the CLI or Python SDK.
Add the cloud account you want to run in.
Send a Docker image, GPU target, command, and buckets.
anycloud starts the VM and runs your container.
Interrupted spot jobs restart with checkpoints restored.
Outputs are uploaded and the VM is terminated.
Five minutes from install to running hundreds of experiments.