Launching today

Locant
Point at any Mac app. Your coding agent gets the element.
1 follower
Point at any Mac app. Your coding agent gets the element.
1 follower
Asking a coding agent to change something on screen means describing it in words or pasting a screenshot, and the agent has to guess which code you mean. Locant lets you point instead: press Control twice, click the element in any Mac app, the iOS Simulator or a web page, and type the change. Your agent gets the element itself: its name, ID and a crop, plus the React component and source line on your dev server. Works with Claude Code, Cursor and Codex. Free and open source.








Free
Launch Team / Built With

AWS Activate CreditsCredits that fuel your startup’s next milestone
Promoted

Hi Product Hunt, I'm Malik, a designer. I built Locant because I kept describing UI to coding agents in words, and they kept editing the wrong thing.
Every Mac app already has an accessibility tree. Locant reads it. Press Control twice, click one element, type one sentence, and your agent gets the element's role, label, identifier, frame and a crop. For apps you build in Xcode, it also names the folder the running build came from.
Agent apps now have pickers inside their own browser panes. Locant works outside them: native Mac apps, the iOS Simulator, any browser, and any agent (Claude Code, Cursor and Codex are tested). Since 1.3, on a page your dev server serves in Chrome, Dia, Edge or Brave, the capture also names the React component and its source line.
I measured it: 40 runs where the app on screen came from another git worktree. Given a screenshot, the agent changed the wrong worktree's code in 14 of 20 runs. Given a Locant capture, 0 of 20. The report is on GitHub.
Free and MIT. No account, no telemetry; captures stay on your Mac. macOS 15 or later. You can try it in your browser on the site before you install.
One question for teams: where do your designers look at builds — a phone, the Simulator, or a Mac app?