OpenBrowser
openbrowser.me
- Windows
- macOS
- Linux
- API
- Self-hosted
What OpenBrowser is
OpenBrowser is a free, open-source framework for browser automation. An LLM writes Python code that navigates web pages, interacts with them, and extracts information. The framework communicates directly through the Chrome DevTools Protocol and supports screenshot analysis for visual understanding. Its MCP server exposes one execute_code tool that runs Python in a persistent namespace. Listed LLM providers include OpenAI, Anthropic, Google, Groq, AWS Bedrock, Azure, Ollama, DeepSeek, Cerebras, OpenRouter, OCI, and others. Compatibility is listed for MCP clients and tools such as Claude Code, Cursor, Windsurf, VS Code, n8n, Cline, and Roo Code. Users can record and replay browser sessions, then export them to Jupyter notebooks or API crawler code. Example uses include research, scraping, form filling, end-to-end testing, page analysis, accessibility audits, and file downloads. Installation instructions cover Windows, macOS, and Linux, with Docker, Kubernetes, and cloud deployment. A hosted version is offered through a waitlist and is described with a web UI, VNC streaming, saved logins, scheduled workflows, and email notifications.
Who it suits
OpenBrowser suits developers and automation users who want LLM-driven browser control through Python and MCP. It also fits teams considering recorded workflows, notebook or crawler exports, or the described hosted platform.
Good to know, and what to watch
- Uses direct Chrome DevTools Protocol communication
- Runs Python through an MCP execute_code tool
- Supports many listed LLM providers and MCP clients
- Records, replays, and exports browser sessions
- Supports Windows, macOS, Linux, Docker, Kubernetes, and cloud deployment
- Hosted version requires joining a waitlist
- MCP allowed-domain whitelist is unset by default
- Hosted features are described rather than generally available
The verdict
OpenBrowser combines code-driven browser automation, visual page analysis, and MCP access with broad provider compatibility. The waitlisted hosted offering and default domain-whitelist setting deserve attention during deployment planning.
What the maker publishes
- Agent interface
- Its MCP server exposes one execute_code tool that runs Python in a persistent namespace for browser automation. source
- Automation
- Its MCP server exposes browser automation through one execute_code tool that runs Python code in a persistent namespace. source
- Browser compatibility
- The quickstart says OpenBrowser detects installed Chromium-based browsers including Chrome, Edge, Brave, and Chromium, and can install Chromium automatically on first run if none is found and uvx is available. source
- Browser control
- It uses direct Chrome DevTools Protocol communication and supports screenshot analysis for visual understanding of pages. source
- Browser control
- It uses direct Chrome DevTools Protocol communication and supports screenshot analysis for visual understanding of pages. source
- Deployment
- The homepage says the product supports Docker, Kubernetes, and cloud deployment. source
- Deployment
- The production guide describes persistent browser profiles, headless production use, allowed-domain restrictions, proxies, and connecting to remote browsers through CDP. source
- Developer integrations
- The maker lists compatibility with Claude Desktop, Claude Code, Codex, OpenCode, other MCP clients, and n8n workflows. source
- Hosted features
- The repository describes the hosted platform as including a web UI, real-time VNC streaming, saved logins, scheduled workflows, and email notifications. source
- Hosted features
- The repository README says its hosted version includes a cloud dashboard for saved cookies and scheduled tasks; the homepage invites visitors to join a waitlist for hosted early access. source
- Hosted security
- The repository README describes hosted saved logins as KMS-encrypted and scheduled workflows as using EventBridge and SQS. source
- Hosted version
- The homepage invites users to join a waitlist for early access to the hosted version. source
- Installation platforms
- The repository provides installation instructions for Windows, macOS, and Linux. source
- Installation platforms
- The maker provides installation instructions for macOS, Linux, and Windows, including shell or PowerShell installers and Python package installation. source
- Integrations
- The maker lists compatibility with MCP clients and names Claude Code, Cursor, Windsurf, VS Code, n8n, Cline, and Roo Code. source
- License
- The maker describes OpenBrowser as open source under the MIT license. source
- License
- The homepage states that OpenBrowser is open source under the MIT license. source
- LLM integrations
- The documentation lists integrations including Google Gemini, OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Groq, OCI, Ollama, Qwen, DeepSeek, Novita, and OpenRouter, plus OpenAI-compatible models. source
- LLM providers
- The repository lists support for providers including OpenAI, Anthropic, Google, Groq, AWS Bedrock, Azure, Ollama, DeepSeek, Cerebras, OpenRouter, OCI, and Browser-Use. source
- Product
- OpenBrowser is a framework for browser automation in which an LLM writes Python code to navigate, interact with, and extract information from web pages. source
- Product
- OpenBrowser describes itself as a framework for intelligent browser automation in which an LLM writes Python code in a persistent namespace to navigate, interact with, and extract information from web pages. source
- Security
- The repository says hosted saved logins are KMS-encrypted. source
- Security controls
- The MCP documentation lists an optional comma-separated allowed-domain whitelist, which is unset by default. source
- Support
- The documentation directs users to GitHub Issues and Discord, and lists an email address for enterprise support. source
Checked against the maker's pages on 9 Oct 2026.