# Skyvern — AI Visual Browser Automation Agent

> Automate any website using LLMs and computer vision. No selectors needed — works on sites never seen before. 21K+ stars.

## Install

Save the content below to `.claude/skills/` or append to your `CLAUDE.md`:

# Skyvern — AI Visual Browser Automation Agent

## Quick Use

```bash
pip install skyvern
skyvern quickstart
```

This launches the Skyvern UI at `http://localhost:8080`. Or use the SDK:

```python
from skyvern import Skyvern

skyvern = Skyvern(api_key="YOUR_KEY")

# Navigate and interact with any website
task = skyvern.create_task(
    url="https://store.example.com",
    goal="Add the cheapest laptop to cart and proceed to checkout",
)
result = skyvern.get_task(task.task_id)
print(result.status, result.extracted_data)
```

Playwright SDK (new):

```python
from skyvern import SkyvernPlaywright

async with SkyvernPlaywright() as skyvern:
    page = await skyvern.new_page()
    await page.goto("https://example.com")
    await page.act("Click on the login button")
    data = await page.extract("Get the user profile information")
```

---

## Intro

Skyvern is an AI browser automation platform with 21,000+ GitHub stars that combines LLMs with computer vision to automate any website — even ones it has never seen before. Unlike traditional automation tools that rely on brittle CSS selectors and XPaths, Skyvern visually understands web pages and plans actions like a human would. It offers a Python/TypeScript SDK (as a Playwright extension), a no-code workflow builder UI, and a managed cloud with anti-bot handling and proxy rotation. Used in production by companies automating procurement, data entry, and web research at scale.

Works with: Any website, OpenAI GPT-4o, Anthropic Claude, Playwright. Best for teams automating complex web workflows that break with traditional selectors. Setup time: under 5 minutes.

---

## Skyvern Architecture & Capabilities

### Three Interaction Modes

| Mode | How It Works | Best For |
|------|-------------|----------|
| **AI Mode** | Pure natural language — LLM + vision decides what to do | Unknown/dynamic websites |
| **Selector Mode** | Traditional Playwright CSS/XPath selectors | Known, stable pages |
| **AI-Fallback** | Tries selector first, falls back to AI if it fails | Production reliability |

### Core AI Commands (Playwright SDK)

```python
# Act — Perform an action described in natural language
await page.act("Fill in the email field with test@example.com")
await page.act("Click the submit button")
await page.act("Select 'Express' from the shipping dropdown")

# Extract — Pull structured data from the page
pricing = await page.extract("Get all pricing plans with features")

# Validate — Check if a condition is met
is_logged_in = await page.validate("Is the user currently logged in?")

# Prompt — Ask the LLM a question about the page
answer = await page.prompt("What payment methods does this site accept?")
```

### Visual Understanding

Skyvern uses computer vision to:
- Identify interactive elements (buttons, forms, dropdowns) by appearance
- Read and understand page layout without DOM access
- Handle CAPTCHAs and visual challenges
- Adapt to layout changes automatically

### No-Code Workflow Builder

The UI at `localhost:8080` provides:

```
┌─────────────────────────────────────┐
│  Workflow: Auto-fill Job Application │
├─────────────────────────────────────┤
│  Step 1: Navigate to job posting    │
│  Step 2: Click "Apply Now"          │
│  Step 3: Fill personal info         │
│  Step 4: Upload resume (PDF)        │
│  Step 5: Submit application         │
│  Step 6: Extract confirmation #     │
└─────────────────────────────────────┘
```

Build multi-step workflows visually, then run them on schedule or via API.

### Cloud Features

The managed cloud offering adds:
- **Anti-bot detection** — Bypasses common bot protection systems
- **Proxy network** — Automatic IP rotation across regions
- **CAPTCHA solving** — AI-powered CAPTCHA completion
- **Parallel execution** — Run hundreds of browser sessions simultaneously
- **Session recording** — Full video replay of every automation run

### Real-World Use Cases

| Use Case | Example |
|----------|---------|
| **Procurement** | Auto-fill purchase orders across vendor portals |
| **Insurance** | Fill out quote forms on multiple carrier sites |
| **HR** | Submit job applications across multiple boards |
| **Research** | Extract data from government and financial sites |
| **Testing** | E2E testing that adapts to UI changes |

---

## FAQ

**Q: What is Skyvern?**
A: Skyvern is an AI browser automation platform with 21,000+ GitHub stars that uses LLMs and computer vision to automate any website without brittle selectors, offering a Python/TypeScript SDK, no-code UI, and managed cloud.

**Q: How is Skyvern different from Stagehand or Browser Use?**
A: Skyvern uniquely combines computer vision with LLMs for visual page understanding, offers a no-code workflow builder, and provides a managed cloud with anti-bot handling. Stagehand is a TypeScript-first library; Browser Use is a Python agent framework. Skyvern is best for enterprise automation at scale.

**Q: Is Skyvern free?**
A: The open-source version (AGPL-3.0) is free to self-host. Skyvern Cloud offers a free tier with paid plans for production scale.

---

## Source & Thanks

> Created by [Skyvern-AI](https://github.com/skyvern-ai). Licensed under AGPL-3.0.
>
> [skyvern](https://github.com/skyvern-ai/skyvern) — ⭐ 21,000+

Thanks to the Skyvern team for bringing visual AI understanding to browser automation.


---
Source: https://tokrepo.com/en/workflows/skyvern-ai-visual-browser-automation-agent-6da285ea
Author: Agent Toolkit