Show HN: Let your AI agents paint big arrows, boxes and text on your screen
Points and comments are a snapshot, not live.
A macOS CLI tool lets AI agents draw arrows, boxes, and text on screen to guide humans through clicks.
The `bigarrow` tool draws a transparent overlay arrow pointing at UI elements, with customizable style, color, shape, and text. It needs no macOS permissions, runs as a single Swift binary, and arrows auto-remove after a timeout or when the agent process ends. Commands include `point` (single arrow), `start`/`stop` (persistent arrows), and support for targeting by element label, coordinate, or accessibility identifier. Built for agents that hit steps only a human can do: permission prompts, 2FA, payments, or guided tutorials. Licensed MIT, installable via Homebrew.
What commenters are saying
Commenters split into camps. Some see genuine value: for guiding parents through settings, for learning complex apps like Blender, and for documentation where "click X" with an arrow beats text descriptions. Others are skeptical: if the human is just a "sudo-clicker" for the bot, why not give it root access? Several note the tool's potential as a "tutorial mode" or for agents to visually annotate parts of the UI. One commenter found irony in criticizing the project's AI-generated tone while praising its utility.