Your AI-powered browser automation assistant. Control the web with natural language.
0424.mp4
BrowserBee is a privacy first open source Chrome extension that lets you control your browser using natural language. It combines the power of Claude 3.7 Sonnet for instruction parsing and Playwright for robust browser automation.
Since BrowserBee runs entirely within your browser (with the exception of the LLM), it can interact with logged-in websites, like your social media accounts or email, without compromising security or requiring backend infrastructure. This makes it more convenient for personal use than other "browser use" type products out there.
- Control your browser with natural language commands
- Take screenshots and analyze web content
- Navigate between tabs and manage browser windows
- Interact with web elements (click, type, etc.)
- Social Media Assistant: Checks your social media accounts, summarizes notifications and messages, and helps you respond.
- News Curator: Gathers and summarizes the latest headlines from your preferred news sources and blogs, giving you a quick, personalized briefing.
- Personal Assistant: Helps with everyday tasks like reading and sending emails and messages, booking flights, finding products, and more.
- Research Assistant: Assists with deep dives into topics like companies, job listings, market trends, and academic publications by gathering and organizing information.
- Knowledge Bookmarking & Summarization: Quickly summarizes articles, extracts key information, and saves useful insights for later reference.
Please refer to ROADMAP.md for an up to date list of features we're aiming to add to BrowserBee.
- Support for saving and replaying sessions (macros)
- Ability to memorize key information as needed (in your local Chrome instance using IndexedDB)
- Scheduled task execution (e.g. check news and social media every morning)
- Ability to learn from your local and personal usage, i.e. memorize and recall optimal tool use sequences for a given website/task
- Support for additional LLMs
If you're interested in contributing to build any of these features or to improve BrowserBee in any way, please head over to CONTRIBUTING.md.
- Clone this repository
- Install dependencies with
npm install - Build the extension with
npm run build - Load the extension in Chrome:
- Go to
chrome://extensions/ - Enable "Developer mode"
- Click "Load unpacked" and select the
distdirectory - Set your Anthropic API key in the options page that pops up
- Go to
- Click the BrowserBee icon in your Chrome toolbar to open the side panel
- Type your instruction (e.g., "Go to Google, search for Cicero, and click the first result")
- Click Execute – and watch BrowserBee go to work 🐝
npm run dev: Start the development servernpm run build: Build the extension for production
BrowserBee uses a modular agent architecture with three key modules:
- Agent Module – Processes your instructions and maps them to browser actions
- Background Module – Manages tab control, messaging, and task streaming
- UI Module – Provides a clean sidebar interface for interaction and configuration
Each module is designed with separation of concerns in mind, making the codebase maintainable and extensible.
For detailed information about the architecture, component structure, and code organization, please see ARCHITECTURE.md.
- Works best in a single Chrome window
- Some websites may block browser automation
BrowserBee is built using these amazing open source projects:
- Cline enabled us to vibe-code the first version of BrowserBee and inspired us to build a "Cline for the web" in the first place
- playwright-crx by @ruifigueira for in-browser use of Playwright
- playwright-mcp for the browser tool implementations