All-in-One Sandbox
A self-hosted AI agent sandbox combines browser automation, shell access, files, VSCode and Jupyter in one Docker container with API and MCP access.
All-in-One Sandbox gives AI agents a place to browse websites, execute code and work with files. It's for developers building agents that need these tools together, with a shared filesystem instead of separate environments for each task.
- Shared files: Browser downloads are available to shell commands, file tools and VSCode within the same container.
- Browser automation: Agents can control the browser through Chrome DevTools Protocol or MCP, while VNC provides visual access to the browser session.
- Code execution: Python and Node.js run in isolated sandboxes, alongside terminal access for command execution.
- Development workspace: Browser-based VSCode, Jupyter notebooks and persistent terminals support interactive work, with port forwarding for web app previews.
- MCP tools: Included servers handle browser actions, file operations, shell commands and document conversion through Markitdown, including text and image extraction.
The open-source project uses the Apache 2.0 license and runs as a self-hosted Docker container. Its API and SDKs let agent applications access the environment, with SDKs for Python, TypeScript/JavaScript and Go. The browser, files and development tools run together on the machine or server hosting that container.
Logotyper
