---
name: unbrowser-intelligence
description: "Lightweight Unbrowser intelligence for AI browser tools. Try this FIRST before opening browser windows - get content via cache, API discovery, or stealth fetch. Use alongside Claude-in-Chrome, browser-use, or other LLM browsers."
---

# Unbrowser Intelligence

**Try Unbrowser FIRST before using browser automation.**

This lightweight profile provides just 2 tools designed to work alongside AI browser tools like Claude-in-Chrome, OpenAI Atlas, browser-use, and other LLM-driven browsers.

## Why Use This?

If Unbrowser can get the data via cache, API discovery, or stealth fetch, you save time and users don't see browser windows opening.

**Workflow:**
1. Call `unbrowser_check(url)` - See if Unbrowser can handle it
2. If `canFetchDirectly: true`, call `unbrowser_get(url)` - Get content instantly
3. If successful, use the markdown/structured data (no browser needed!)
4. If unavailable, use `browserHints` for effective automation

## Tools Included (2)

- `unbrowser_get` - Fetch content without browser automation
- `unbrowser_check` - Check what we know before browsing

**Token usage: ~800 tokens**

## Installation

```bash
npx unbrowser init --profile=intelligence
```

Or configure manually:

```json
{
  "mcpServers": {
    "unbrowser-intelligence": {
      "command": "npx",
      "args": ["unbrowser", "--intelligence"]
    }
  }
}
```

## Example Usage

```
Check if we can get HackerNews without a browser:
> unbrowser_check https://news.ycombinator.com

Response: canFetchDirectly: true (has learned patterns)

Get the content:
> unbrowser_get https://news.ycombinator.com

Response: { success: true, source: "api", fetchDuration: 413ms }
(Used HackerNews Firebase API - 10x faster than browser!)
```

## Performance

| Method | Time |
|--------|------|
| Browser automation | 2-5 seconds |
| Unbrowser intelligence | 50-500ms |

**Speedup: 5-50x faster**