Open Browser is an open-source TypeScript framework for autonomous web browsing. The current project centers on letting agents click, type, navigate, and extract data on live websites, with built-in support for OpenAI, Anthropic, and Google models plus a sandbox package for constrained execution.
Features
- Playwright-based browser control for clicking, typing, navigation, extraction, and screenshots
- Natural-language agent runs that translate tasks into multi-step browser actions
- Interactive REPL for debugging and exploring live browser sessions
- Multi-model support through OpenAI, Anthropic, and Google providers
- Optional sandboxed execution with resource limits, monitoring, and allowed-domain controls
So agents can…
- Complete web tasks on dynamic sites without writing a full browser script for each flow
- Drop into an interactive browser session to inspect or recover a failing task
- Run browser automation inside constrained environments with timeouts and domain restrictions
Comments