Visual context for AI agents

Point at your UI. Paste a perfect prompt.

Point2Prompt is a bookmarklet that lets you select real UI elements, attach change intent, and copy a structured prompt for Claude Code, Codex, Cursor, or a GitHub issue.

Point2Prompt

Drag the green button to your bookmarks bar, or click it right here to test on this page. Then open the recording demo.

Localhost:3000

Tell the agent exactly where to work.

Stop writing "the small text under the second card." Select it, choose the intent, and ship the context your coding agent needs.

Change brief Codex

h1.hero-title

Intent: Improve visual hierarchy

Includes DOM selector, text, bounding box, computed styles, React source hints, and your instruction.

The missing pointing layer for AI coding.

Coding agents are good at changing files. Humans are good at seeing what feels wrong. Point2Prompt connects those two worlds.

Select real UI, not vague descriptions

Click any element, shift-click for multi-select, or walk the DOM with keyboard navigation.

Attach intent before copying

Pick presets like copy, spacing, hierarchy, contrast, responsive layout, or write your own instruction.

Export for your workflow

Switch between formats for Claude Code, Codex, Cursor, GitHub issues, or plain Markdown.

Everything stays client-side

The bookmarklet injects local UI into the page. No account, no backend, no telemetry, no uploaded DOM.

Claude Code Codex Cursor GitHub Issue Markdown