Add AI Search to Your Agent with AnyCrawler Skills

AI agents need more than a browser. They need a reliable way to discover sources, read them cleanly, and turn that evidence into useful answers.

AnyCrawler now provides two agent skills for that workflow:

• $anycrawler-search for public search across web pages, images, news, videos, and scholar results

• $anycrawler-read for crawling pages, extracting markdown, and capturing screenshots

Together, they give agents a simple search-to-read pipeline.

A typical flow looks like this:

Use $anycrawler-search to find recent news about open-source AI agents. Then use $anycrawler-read to crawl the top sources and summarize them with links.

$anycrawler-search supports:

POST /v1/search/page POST /v1/search/images POST /v1/search/news POST /v1/search/videos POST /v1/search/scholar

$anycrawler-read supports:

POST /v1/crawl/page POST /v1/crawl/screenshot

This lets an agent first discover relevant results, then convert selected pages into compact markdown before reasoning. That is cleaner and usually cheaper for LLMs than raw HTML or browser dumps.

Install:

bash

git clone https://github.com/AnyCrawler-com/AnyCrawler-Skill.git mkdir -p ~/.codex/skills cp -R AnyCrawler-Skill/skills/anycrawler-read ~/.codex/skills/anycrawler-read cp -R AnyCrawler-Skill/skills/anycrawler-search ~/.codex/skills/anycrawler-search export ANYCRAWLER_API_KEY="sk-your-key"

After restarting the agent session, you can ask:

Use $anycrawler-search to search scholar results for RAG benchmarks, then use $anycrawler-read to crawl the most relevant sources and make a comparison table.

AnyCrawler Skills make AI search practical: discover sources, read pages as markdown, and answer with evidence.