OpenSandbox and agent-sandbox
About OpenSandbox
alibaba/OpenSandbox
OpenSandbox is a general-purpose sandbox platform for AI applications, offering multi-language SDKs, unified sandbox APIs, and Docker/Kubernetes runtimes for scenarios like Coding Agents, GUI Agents, Agent Evaluation, AI Code Execution, and RL Training.
This platform allows AI developers to safely test and run their AI applications, like coding agents or those that automate web browsers, in isolated environments. You input your AI code or agent, and it provides a secure, controlled space for execution and evaluation without affecting your main systems. It's designed for AI engineers and researchers who build and deploy intelligent agents and applications.
About agent-sandbox
Parassharmaa/agent-sandbox
A sandboxed execution environment for AI agents via WASM
This project provides a secure, isolated environment where you can safely run AI agents that need to execute command-line tools, shell scripts, or JavaScript code. It takes your agent's instructions (like 'run grep' or 'fetch data') and executes them in a controlled space, returning the results. This is for developers building AI agents who need to ensure their agents can interact with system commands and network resources without risking the host system.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work