GitHub radar
ax — a web scraper designed for AI agents
A tool for AI agents to use instead of curl to get web pages, parse structure, and stay within LLM context window limits — streamlining the fetch + understand + extract cycle that all agents need to do over and over again.
ax is a CLI tool that acts like ‘curl for the AI era’. Fetch webpages and get back structured, token efficient content that’s tailored for LLM agents instead of just plain old HTML. Find outlines, pinpoint text locations, pull out fields, parse tables, all in one go. Written in Bun and just one runtime dependency. Runs entirely locally, no need for any cloud APIs or API keys.
Why a vibe-coder should care
Web page fetching is a key part of how AI code agents get information, but the extra tokens from the HTML can be a bottleneck and expensive. ax dramatically reduces the overhead involved, as demonstrated by these benchmark results from real Claude Code sessions (see the 23–67% token savings). That means you can run your agents faster and cheaper!
▌ More finds