Use cases · Sandboxes · code execution

Sandboxes · code execution

All AI & Agents Sandboxes · code execution High-memory · Pro (RAG, agent fleets, private LLM)Bots Websites, mail & services (dedicated IP)Self-hosting Privacy & Anti-censorship

Sandbox for AI agents: run untrusted code in a microVM that starts in a second

Your agent writes code it has never run before. A Firecracker microVM sandbox gives that code a disposable Linux box with no access to your machine, your keys or your network — started in about a second, billed per second.

→ sandboxes

Sandbox for CI: isolated test runs for under a cent each

Run tests for code you don't trust — a pull request from a stranger, a generated patch — in a fresh microVM that's deleted after the run. Per-second billing, Python/TS SDK, honest about the concurrency limits.

→ sandboxes

Sandbox for LLM evaluation: run 1,000 model-written samples for about a cent

Scoring a model on coding tasks means executing thousands of programs nobody wrote by hand. Run them in disposable microVM sandboxes instead of on your laptop — per-second billing, with an honest note on throughput and CPU limits.

→ sandboxes

Sandbox for code review and PR checks: prove a patch works before you approve it

Reading a diff tells you what a patch claims. Running it tells you whether it's true. Check out a pull request in a disposable microVM, reproduce the bug before and after, run the tests, and review with evidence — for about half a cent per pass.

→ sandboxes