Source checked · Build

OpenSandbox

A runtime for isolated AI-agent execution.

Best for: Infrastructure teams evaluating a managed sandbox service with explicit command, file, ingress, egress, and credential policies.

Limitations: Treating a Docker or Kubernetes sandbox as a complete trust boundary without reviewing its server, host, network, and secret controls.

Public use

12,004 GitHub stars · 1,003 forks

Activity

Release java/code-interpreter/v1.0.16 on 13 Jul 2026 · repository updated 14 Jul 2026

Community

63 open issues · Discussions available

Checked

14 Jul 2026

Source facts only. Not tested or recommended by Agent Blocks.