OpenSandbox Examples

Examples for common OpenSandbox use cases. Each subdirectory contains runnable code and documentation.

Integrations / Sandboxes

aio-sandbox: All-in-one sandbox setup using OpenSandbox SDK and agent-sandbox
agent-sandbox: Create a kubernetes-sigs/agent-sandbox instance and run a command
code-interpreter: Code Interpreter SDK singleton example
host-volume-mount: Mount host directories into sandboxes (read-write, read-only, subpath)
rl-training: Reinforcement learning training loop inside a sandbox
claude-code: Call Claude (Anthropic) API/CLI within the sandbox
iflow-cli: CLI invocation template for iFlow/custom HTTP LLM services
gemini-cli: Call Google Gemini within the sandbox
codex-cli: Call OpenAI/Codex-like models within the sandbox
langgraph: LangGraph agent orchestrating sandbox lifecycle + tools
google-adk: Google ADK agent calling OpenSandbox tools
openclaw: Run an OpenClaw Gateway inside a sandbox
desktop: Launch VNC desktop (Xvfb + x11vnc) for VNC client connections
playwright: Launch headless browser (Playwright + Chromium) to scrape web content
vscode: Launch code-server (VS Code Web) to provide browser access
chrome: Launch headless Chromium with DevTools port exposed for remote debugging

Set basic environment variables (e.g., export SANDBOX_DOMAIN=..., export SANDBOX_API_KEY=...)
Add provider-specific variables as needed (e.g., ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY, IFLOW_API_KEY/IFLOW_ENDPOINT, etc.; model selection is optional)
Navigate to the example directory and install dependencies: pip install -r requirements.txt (or refer to the Dockerfile in the directory)
Then execute: python main.py
To run in a container, build and run using the Dockerfile in the directory
Summary: First set required environment variables via export, then run python main.py in the corresponding directory, or build/run the Docker image for that directory.

此页内容来自仓库源文件：examples/README.md