I've been working on integrating AI sandboxes for our agents to run code securely, and kept facing issues with varying API surfaces which caused a lot of bottlenecks when we needed to quickly pivot to other providers for features, pricing, compliance, cost, or other reasons.
Introducing Sandboxer - one client surface for remote sandboxes!
You can open a box, run commands, manage files, and tear down the same way in Go, Python, and TypeScript, whether you’re on E2B, Daytona, Blaxel, Runloop, Fly.io, or Docker on your machine.
You don’t need another opinionated platform in the path - you need one mental model and the freedom to swap hosts when requirements change.
Here's where Sandboxer comes in: - Unified API across languages for the workflows teams actually repeat: lifecycle + exec + filesystem. - No Sandboxer service in the request path, your app talks directly to each provider (or the local Docker flow where applicable). - Your credentials stay in your boundary.
Ship integrations once, keep optionality across vendors, reduce glue code and review surface area.
There are 75+ examples across various providers and SDKs in the repository.
Really appreciate your support!