Agent Sandbox
A secure code execution tool for AI agents.
What is Agent Sandbox?
Agent Sandbox is a trusted runtime that lets AI agents run Python and shell commands inside a private, sandboxed workspace via a single API for code, sessions, and artifacts. It offers a secure, sandboxed environment by default, automatic dependency installation, and data flow capabilities that let you upload inputs, execute tasks, and retrieve outputs such as charts and reports, enabling integration with any agent framework either client-side or server-side.
Key features of Agent Sandbox
Secure runtime for untrusted code
One API for code, sessions, and artifacts
Sandboxed by default
Run Python and shell commands in a safe workspace
Pros of Agent Sandbox
Secure, sandboxed execution for untrusted code
