AI Agent
The AI Browser Agent is an autonomous agent that controls a cloud browser using natural language instructions. It can navigate pages, fill forms, click buttons, extract data, and solve CAPTCHAs - all without writing any selectors or automation code.
How It Works
- You describe a task in natural language (e.g., "Go to Amazon and find the top 5 wireless headphones under $50")
- The agent launches a cloud browser session
- At each step, it takes a screenshot, analyzes the page, and decides the next action
- Actions include clicking, typing, scrolling, navigating, and extracting data
- The agent returns structured results when the task is complete
Key Features
- Natural language tasks - Describe what you want, not how to do it
- Visual understanding - Uses vision models to interpret page content
- Self-healing - Automatically retries and adapts when elements change
- CAPTCHA solving - Optional automatic CAPTCHA solving (reCAPTCHA, hCaptcha, Turnstile)
- Multiple AI models - Choose from budget to ultra-tier models
- Convert to workflow - Turn successful agent runs into reusable workflows
- Real-time streaming - Watch the agent work in real time via the dashboard
Available Models
| Tier | Credits/Iteration | Models |
|---|---|---|
| Budget | 2 | Gemini 3 Flash, GPT-5 Nano, Qwen3 VL Flash, Grok 4 Fast |
| Standard | 5 | GPT-5 Mini, Claude 4.5 Haiku, Qwen3 VL Plus, MiniMax M2.5, Kimi K2.5 |
| Performance | 10 | GPT-5, Gemini 3.1 Pro, GPT-5.2, Claude 4.5 Sonnet |
| Premium | 25 | Claude Sonnet 4.6, GPT-5.4, Claude Opus 4.6, Claude Opus 4.5, Grok 4 |
| Ultra | 50 | GPT-5.4 Pro, Claude Opus 4.1 |
💡Pricing
Total cost = 5 (base session) + (maxIterations x creditsPerIteration) + 3 (if CAPTCHA solver enabled). See Credits & Billing.
Dashboard Usage
The easiest way to use the AI Agent is through the Scrayle Dashboard:
- Navigate to the AI Agent page
- Enter your task description
- Optionally set a start URL, model, iteration limit, and CAPTCHA solving
- Click Run and watch the agent work in real time
- Review extracted data and optionally convert the run to a workflow
Quick Links
- Creating Sessions - API and SDK usage
- Convert to Workflow - Reuse successful runs
- AI Agent API Reference