Use Cases

CodeSandbox SDK has been designed to help you build and launch any projects that require:

• An isolated code interpreter.
• Running multiple development environments at scale.
• Access to a sandboxed environment.

There are hundreds of use cases that can benefit from these three features. Below are some of the most interesting ones we've come across so far:

AI Agents

Agentic workflows are a fast-emerging use case that benefit immensely from CodeSandbox SDK.

Using the SDK, You can give each agent a sandbox to resolve user prompts, or you can create autonomous agents running on sandboxes.

Doing this ensures they are securely isolated from your system, while still allowing them to interact with each other as required.

You can also run multiple agents in parallel, if you want to avoid interaction entirely.

Using the forking mechanism, you can also A/B test different agents.

Development environments

You can use the SDK to programmatically provision a sandbox for anybody who needs an environment to code. Allowing you to run multiple development environments in parallel without them interfering with each other, all accessible through the browser.

With the SDK, you can essentially create your own CodeSandbox for your team/company/school!

Code Interpretation

CodeSandbox SDK allows programmatically spinning up sandboxes (VMs) to execute code in. Providing an ideal environment to interpret code, for both for simple and complex tasks.

As these sandboxes exist in total isolation, you can run untrusted code without worrying about it affecting your system.

Many early adopters, such as Blackbox AI, are leveraging the power of CodeSandbox SDK to bring code interpretation capabilities to their generative AI applications. We highly recommend following the guide by Together AI (opens in a new tab) to get CodeSandbox SDK running seamlessly with any of the leading open-source LLMs.

CI/CD

With the SDK, you can set up your own CI/CD implementation to run tests inside a sandbox and hibernate the sandbox when the tests are done. With our snapshotting technology, you can quickly pick up where you left off if you start the sandbox again to re-run your tests or evaluate the results.

You can also use the SDK to create 1 sandbox that runs your CI/CD preparation, and then clone it multiple times to run your tests in parallel.