Whiteboard
A desktop canvas for reviewing what your coding agent actually built
Emerged finds new products worth a look, daily — what’s new today
Answer me with HTML
Your coding agent hands back a built page instead of a wall of prose
The model writes only a short Markdown draft and a bundled renderer does the layout, which on the author's own small benchmark costs about an eighth of the tokens of writing the HTML itself.
An agent skill, not an app. Installed into Claude Code, Codex, Cursor, OpenCode or Pi, it decides when a question deserves a page, writes a brief Markdown draft, then runs a bundled CLI that handles layout, diagrams and styling, saving one offline HTML file. It needs Node 20 and an agent that can run shell commands; install via plugin marketplace, npx skills or a folder copy. The project site runs the real renderer in a playground. MIT, 2.3k stars, 134 commits. Pages pile up and need clearing by hand, and the speed figures come from nine pages.
More like this
Sphinx
A pre-push gate that asks what the agent's change does before it ships
ZCode
Zhipu's open-source workbench for running AI agents on a codebase
Modern Web Guidance
Browser-vendor guides that teach coding agents today's built-in web APIs
Thrum
A local message bus that lets your coding agents review each other's work
Narrative Deck
An agent skill that turns raw notes into a nine-beat web article