
Browser Use Chat UI Example
https://github.com/browser-use/chat-ui-example- Category
- AI Agents
- Rank
- No. 1593Tools index
Previous survey · No. 1599 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- browser-use
- GitHub
- 21 stars
- Date
About
Reference chat application demonstrating the Browser Use v3 SDK with a browser-controlling AI agent.
What it does
Users describe a web task, then watch the agent act through an embedded browser while progress messages stream into a chat. Sessions stay open for follow-up requests, can use saved profiles or workspace context, support country proxies, and produce downloadable browser recordings.
Why it's ranked here
This is a focused, inspectable starting point for teams building a hosted browser-agent interface. It covers the full interaction loop, including session creation, streamed progress, stopping work, follow-ups, live viewing, and recordings. Its dependence on Browser Use Cloud makes it less useful as a general browser automation foundation.
What's good
The API key remains server-side. Streaming avoids manual polling during tasks, while the interface turns raw messages and tool results into readable steps. Users can choose models, reuse authenticated profiles, attach workspace context, select proxy countries, stop active work, and continue within persistent sessions.
Tradeoffs
Running it requires a Browser Use API key and the hosted SDK. The live browser panel is desktop-only. Profile listing still uses the older SDK client because that capability is unavailable on the newer client. The supplied project structure shows linting but no test suite or test script.
How to use it well
Use it when prototyping or teaching the interaction model for a browser agent inside a React application. Fork the interface, study the session and streaming flow, then adapt the presentation to your product. Choose something else for self-hosted, always-on operation or automation independent of Browser Use Cloud.
Technical notes+
The private npm application uses Next.js 15, React 19, TypeScript, Tailwind CSS, and browser-use-sdk. src/lib/api.ts creates both v3 and v2 clients from the server-only BROWSER_USE_API_KEY. src/lib/actions.ts creates persistent, recordable sessions through v3 but lists profiles through v2. src/context/session-context.tsx consumes server-sent events from /api/stream/[sessionId], deduplicates messages, tracks task state, and retrieves recordings after terminal states. src/lib/message-converter.ts parses OpenAI-format message data in two passes, associates tool results by call ID, hides internal tools, and groups output into turns and steps. src/components/browser-panel.tsx embeds the session live URL in an iframe.
Observed
- License
- MIT
- Primary language
- TypeScript
- Packaging
- Private npm application installed with npm
- Interface
- Next.js web chat interface backed by the Browser Use Cloud SDK
- Runtime requirement
- Requires a server-side Browser Use API key
- Deployment surface
- Includes local development, production build, production start, and Vercel deployment instructions
- Testing structure
- No test directory or test script appears in the supplied repository text
Read from README.md, package.json, src/lib/api.ts, src/app/page.tsx, src/lib/types.ts, src/app/layout.tsx, src/lib/actions.ts, src/lib/countries.ts, src/lib/tool-labels.ts, src/components/markdown.tsx, src/lib/message-converter.ts, src/components/chat-input.tsx, src/components/step-section.tsx, src/context/session-context.tsx, src/components/browser-panel.tsx.
What it can do
Control web browser through AI agent
Natural language commands → Automated browser actions
Navigate to websites
Website URLs or descriptions → Loaded web pages
Fill out web forms
Form field requirements and data → Completed web forms
Extract data from web pages
Web page content and extraction criteria → Structured data
Click buttons and links
Element descriptions or selectors → Triggered web page interactions
Search for content on web pages
Search terms and target websites → Located content or search results
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.