What Is Leap0?
Leap0 is secure infrastructure for running AI-generated code inside isolated microVM sandboxes with built-in APIs for files, git, processes, and telemetry.
Steven Passynkov
AI agents are great at writing code, but code generation is the easy part. The hard part is executing that code safely and reliably while giving the agent the proper tools to do real work.
That is exactly what Leap0 is for - securing AI-generated code execution.
Leap0 is secure infrastructure for running AI-generated code. You can create a sandbox from a template, start it in around 200ms, invoke it over HTTP, SSE, or WebSocket, interact with files and git, and stream telemetry.
Why Leap0 exists
Modern agents do not just answer questions. They run commands, clone repositories, install dependencies, and ship changes.
If that execution happens directly on your host machine or production environment, one bad command can destroy data, leak secrets, or take down systems.
Leap0 solves this by isolating agent execution in disposable environments built for untrusted code.
I built Leap0 after seeing teams struggle to give AI agents real execution power without exposing their infrastructure to unacceptable risk.
Sandbox capabilities
Leap0 provides agents with:
- Run one-shot or streaming commands
- Manage files and directories
- Clone repositories and use git workflows
- Open persistent terminal sessions
- Access running apps over HTTP, SSE, or WebSocket
- Read logs, metrics, and lifecycle telemetry
This means your agent can execute real software tasks end-to-end, without direct access to your laptop or core infrastructure.
Security model
Leap0 uses Firecracker microVMs with Jailer for process and filesystem isolation, so each sandbox has strong boundaries and an independent failure domain. Sandboxes are short-lived, isolated, and easy to tear down after each run.
Who Leap0 is for
Leap0 is for AI agents that need to run code in a safe execution environment without having to compromise on capability. It gives those agents isolated execution environments instead of direct access to trusted systems. This lets teams ship autonomous workflows with stronger security and predictable operational boundaries.
Public preview
Leap0 is free during public preview while we finalize the API.
Public preview means the platform is usable today, but still evolving quickly:
- Breaking changes may happen as we improve the platform.
- Short periods of downtime may occur.
What’s next
Over the next releases, we are focused on making the API more stable, expanding sandbox capabilities, and improving reliability for production agent workloads. To see more, see our roadmap.
Getting started
If you are new to Leap0, start with the quickstart in the docs.
Ready to give your agents a safe sandbox? Get started. Try it and let us know what you think.