what is ax fr?

ax is a command line tool built for AI coding agents that need to fetch a web page and pull structured information out of it, described by its author as the AI-era version of curl. Instead of an agent piping a raw page download into a throwaway script it has to write and debug each time, ax fetches a page, helps the agent understand what is on it, and extracts organized data in one step, all running locally on the user's own machine with no cloud service involved. A normal web request with curl often prints nothing useful when something goes wrong, leaving an agent guessing. ax instead always reports the status, headers, timing, and final URL of every request, even when the response is empty or an error. It also offers a discovery mode that shows the repeating patterns on a page, such as lists of items, without ever dumping raw HTML into the AI's context window, which would waste a large number of tokens. From there, a user or agent can extract multiple fields from each matching item in a single command, turn HTML tables into clean rows of data, or filter results with a simple expression. The tool is deliberately careful about token usage since it is meant to feed AI agents that have limited context space. Results are capped by default, and a budget option can limit output to an estimated number of tokens, with clear messages telling the agent exactly how to continue reading past a cutoff point. ax is installed with a single shell command or through the Nix package manager, and it can also teach an agent how to use it through a built in skill file. It is built using the Bun JavaScript runtime, compiled into a single binary, and uses the linkedom library for parsing HTML in a standard way. The README includes benchmark comparisons showing lower cost and time when Claude Code agents use ax compared to writing their own scraping scripts.