A sandbox is an isolated, on-demand environment for running, previewing or testing code without touching production. It has its own resources and data, so you can experiment, review a pull request or reproduce a bug freely - and throw the whole environment away when you are done.
How does a sandbox work?
A sandbox is spun up from a known configuration - often automatically by a pipeline when a pull request opens. It gets its own isolated compute, network and (usually) a copy or subset of data, so anything that happens inside it stays inside it.
Because it is disposable, a sandbox can be created for a single branch, feature or experiment and then destroyed once the work is merged or abandoned. This makes it cheap to test risky changes and easy to give every change its own live preview URL.
Why do sandboxes matter?
Sandboxes let teams review changes against a running version of the app instead of guessing from a diff. Reviewers, designers and stakeholders can click through a real environment before anything reaches users.
They also contain risk: because a sandbox is isolated from production, a broken migration or a load test cannot take down the live service or corrupt real customer data.
Example
When a developer opens a pull request, the pipeline builds the branch and deploys it to a fresh sandbox at a unique URL. The team reviews the feature there; once the PR is merged, the sandbox is automatically torn down to free up resources.
Frequently asked questions
What is the difference between a sandbox and a staging environment?
A staging environment is usually a single, long-lived mirror of production. A sandbox is typically ephemeral and per-change - created on demand for a branch or pull request and destroyed afterward.
Is a sandbox the same as production?
No. A sandbox deliberately mirrors production closely but stays isolated, with its own resources and non-production data, so testing in it cannot affect real users.
How is a sandbox created?
Most commonly a CI/CD pipeline creates one automatically from a branch or pull request, deploying the freshly built artifact into the isolated environment.
Suggest a new word or an edit to an existing one. Every submission is reviewed before it goes live.