Secure by Default
Generative AI is spooky good at writing code, but it's not code you can trust. Run your AI workflows safely in a secure sandbox powered by ephemeral Fly Machines.
- Our Kernel-isolation via Firecracker VMs keeps things tight.
- Get isolated code per run, restricted network access, and controlled runtimes.
Fast as Hell
The Machines API spins up ephemeral machines in a few hundred milliseconds, so you’re ready to run code at the drop of an HTTP request. Pay only for the time you're actually up and running.
A Clean Slate for Every Run
The Fly.io platform is built on Fly Machines, API-driven hardware-assisted secure virtual machines. They come up clean on every start, without a trace of whatever ran last on them.
The API We’d Want, If We Were You
Our API is a simple REST/JSON interface that builds Fly Machines out of industry-standard OCI containers. If your Docker or K8s runs it, so will we. You can track running Machines, clone them, time them out, and back them up, straight from your own code. (And they’re open-minded! Fly Machines run generated code in JavaScript, Python, or any other language.)
Whole Apps, Not Just Snippets
With Fly Machines, you don't just get an execution environment; you (safely) get the whole platform. Let your LLM generate an HTTP server, or a SQLite database, files, object storage data, even GPU Machines, so you can put LLMs in your LLMs.
Start now