browser-use

# Browser-Use — AI Browser Automation

## Security & Privacy

- **No credential logging**: Passwords are handled via Browser-Use's `sensitive_data` parameter — the LLM never sees real credentials, only placeholder tokens.
- **User-initiated Chrome connection**: CDP mode (connecting to real Chrome) is opt-in and requires the user to manually launch Chrome with debug flag. The skill never silently connects to running browsers.
- **All packages are open-source**: Dependencies are `browser-use` (38k+ ⭐ on GitHub), `playwright` (by Microsoft), and `langchain-openai` — all widely audited open-source tools.
- **Local execution only**: Scripts run locally on the user's machine. No data is sent to any server except the configured LLM API for step-by-step reasoning.
- **Domain restriction available**: Use `allowed_domains` parameter to restrict which websites the agent can visit.
- **No telemetry**: This skill does not collect, store, or transmit any usage data.

## When to Use Browser-Use vs Built-in Tool

| Scenario | Built-in tool | Browser-Use |
|----------|:-:|:-:|
| Screenshot / click one button | ✅ Free & fast | ❌ Overkill |
| 5+ step workflow (login→navigate→fill→submit) | ❌ Breaks easily | ✅ |
| Anti-bot sites (real Chrome needed) | ❌ | ✅ |
| Batch repetitive operations | ❌ | ✅ |

**Cost**: Browser-Use calls an external LLM per step (costs money + slower). Use built-in tool for simple actions.

## Execution Flow

### 1. Check Environment
```bash
test -d ~/browser-use-env && echo "Installed" || echo "Need install"
```

### 2. First-Time Setup (once only)
```bash
python3 -m venv ~/browser-use-env
source ~/browser-use-env/bin/activate
pip install browser-use playwright langchain-openai
playwright install chromium
```

### 3. Choose Mode
- **Mode A — Built-in Chromium**: For simple automation or when detection doesn't matter. Runs immediately.
- **Mode B — Real Chrome CDP**: For anti-bot sites or when user's login session is needed. Requires user action.

Mode B...