About Fly.io
Fly.io runs apps, sandboxes and databases on hardware-isolated virtual machines. It is built for developers who want their code close to users in several regions.
What it does
Fly Machines are hardware-virtualized containers that start quickly and run when you need them. You can deploy from GitHub or with the fly launch command, and the platform can generate a Dockerfile for many popular frameworks if you do not have one. Sprites are self-contained environments for running untrusted or AI-generated code. They can be checkpointed and restored, which helps when an experiment goes wrong. Storage options include local NVMe, managed Postgres and S3-compatible object storage.
Key features
- Hardware-isolated machines that start fast
- Deploy from GitHub or the fly launch command
- Sandboxes you can checkpoint and restore
- Managed Postgres and Redis options
- Object storage and local NVMe volumes
Who it's for
Developers and teams running web apps, background jobs or AI agents who want per-second billing and control over their machines. Anyone who needs sandboxes for untrusted code, and wants a clear boundary between each one, may also find it useful.
Pricing and good to know
Fly.io bills for the CPU and memory your machines use, down to the second. Pricing details and current rates are on the pricing page. Machines run in the regions you choose, so pick them with your users and your data rules in mind. The CLI and the MCP server are both built for agents to deploy, scale and debug on their own.
Highlights
- Hardware-isolated machines that start fast
- Deploy from GitHub or the fly launch command
- Sandboxes you can checkpoint and restore
- Managed Postgres and Redis options
- Object storage and local NVMe volumes
