Ultra Scraper — AI skill for Claude Code

ultra-scraper

**Agent-native web scraping with automatic stealth escalation.**

Built for AI agents — OpenClaw, Claude Code, Codex, Hermes, and any framework that supports tool/function calling.

[](https://github.com/LeoYeAI/ultra-scraper/releases) [](LICENSE) [](https://myclaw.ai)

🌐 **English** · [中文](docs/README.zh-CN.md) · [Français](docs/README.fr.md) · [Deutsch](docs/README.de.md) · [Русский](docs/README.ru.md) · [日本語](docs/README.ja.md) · [Italiano](docs/README.it.md) · [Español](docs/README.es.md) · [한국어](docs/README.ko.md)

---

Why ultra-scraper?

Most scraping libraries are built for humans writing scripts. ultra-scraper is built for **AI agents making decisions**:

• One function call — scrape(url) handles everything. No config, no tier management.
• Auto-escalation — starts fast (HTTP), escalates to headless → anti-bot → C++ stealth only when needed.
• Native tool schemas — ships with OpenAI, Anthropic, and LangChain tool definitions. One import, zero boilerplate.
• LLM-friendly output — returns structured dicts, not raw HTML soup.

from ultra_scraper import scrape

# Works on 99% of sites out of the box
result = scrape("https://example.com", css=".product")
print(result["elements"])   # [{"text": "...", "html": "...", "attributes": {...}}]
print(result["tier_used"])  # 1 (fast HTTP — no browser needed)

---

Install

pip install scrapling   # core engine (required)

# Optional: C++ source-level stealth for enterprise-grade bot detection
pip install cloakbrowser

Or clone and run setup:

git clone https://github.com/LeoYeAI/ultra-scraper.git
bash scripts/setup.sh

---

Quick Start

Direct Python call (Claude Code / Codex / any agent)

from ultra_scraper import scrape