Browser-Use AI Browser Automation

Problem

Web automation often breaks down when commands are too low-level and the tool lacks page context: manually scripting clicks, inputs, and navigations becomes brittle once the page changes. browser-use shifts the task to an LLM, describing the goal in natural language and letting the model choose the next action from the current page context. It is aimed at dynamic web tasks such as browsing, data extraction, and form filling.

How It Works

The basic flow is to configure an LLM API key, then start a task from the CLI or from JavaScript/TypeScript. It supports providers such as OpenAI, Anthropic, Google Gemini, AWS Bedrock, Azure, Groq, Ollama, DeepSeek, OpenRouter, Mistral, and Cerebras. Compared with playwright-cli, it does not require step-by-step click or input commands; instead, the AI reads the page and selects actions.

Boundaries And Cautions

It fits relatively clear, verifiable automation tasks, not unconstrained high-risk operations. Treat API keys as sensitive data and avoid committing .env; use --allowed-domains to limit where the browser can go; prefer passing sensitive values through sensitive_data.