judge0/judge0

What it solves

Judge0 provides a robust way to run code submitted by users or generated by AI without worrying about security or scalability. It handles compilation and execution of untrusted code in a sandboxed environment, so you can safely run programs you don't trust. This is used for AI agents that need to execute code, coding education platforms, competitive programming judges, and online IDEs.

How it works

The system offers a simple HTTP JSON API and an official Python SDK. You send the source code, language ID, optional standard input, and other settings, and Judge0 compiles and runs the code in a sandbox, returning output, exit status, and memory/time usage. It supports 90+ languages, multi-file programs, custom compiler options, command-line arguments, and limits on time and memory. It can send webhooks (HTTP callbacks) when a submission finishes, and its architecture scales horizontally for high loads.

Who it's for

Developers and organizations building products that need online code execution: AI agent platforms that need to run machine-generated code, e-learning sites with coding exercises, competitive programming platforms, recruitment and candidate assessment tools, online code editors, and IDEs. It's available as a self-hosted open-source service or a fully managed cloud (SaaS).

Highlights

• Sandboxed compilation and execution of untrusted code for security
• Supports 90+ programming languages
• Simple REST API and official Python SDK
• Scalable architecture for high traffic
• Multi-file program support and custom compiler options
• Webhooks for asynchronous notifications
• Self-hostable or available as a managed cloud