Open source, MIT

Intelligent UI Anything

Generative UI for coding agents: turn an answer into a page you can pick from

Agents answer in prose. When the answer is ten models with six attributes each, or twenty findings across three sessions, you scroll, compare in your head and type your choice back. This skill gives the agent five locked page templates instead. The agent writes JSON, a script fills the template, checks it and opens it where you are. You sort, filter and pick, and the pick comes back to the agent as one line.

Install
npx skills add IndusCore-AI/intelligent-ui-anything
Read the code on GitHub

Works in Claude Code, Codex CLI, Cursor, Antigravity, claude.ai. Needs Node.js 18 or later, no packages.

A request in the terminal, the sortable ComfyUI model page it opens, and the pick coming back as: decision: Use model FLUX.2 [klein] 4B
Ask it

Ask the way you would ask a person

  • compare these models and let me pick one
  • lay out the three options for the migration
  • show me what changed, what's risky and what's undecided across today's sessions
  • keep a checklist page updated while you run the release
Why

What Intelligent UI Anything does differently

Five shapes, nothing else

Compare, choose, explain, track, inspect. A question that fits none gets prose.

The decision comes back

Every pick shows the exact line the agent will get. Copy works everywhere; a served page sends it with one click.

Opens where you are

An artifact in Claude Code, your browser in a terminal, a localhost URL over SSH, a path inside a sandbox.

Honest by construction

A page that states a number its data does not hold fails the check and never opens.

Zero dependencies

Node 18 built-ins, no network, one self-contained HTML file per page, light and dark.

Cheaper than hand-written HTML

A 20-item review page took 2,262 output tokens against 4,450 written by hand, measured on 2026-10-25.

How it works

The five shapes

The five shapes
The questionShapeWhat you get
Which of these is best on what matters?compareA sortable, filterable table, an optional bar chart and an optional pick per row
Which one should we do?choose2 to 6 option cards with good and watch-out points, one pick
What does this say, and what changed?explainA long document with contents, anchors and a changed-sections view
How far along is it?trackA checklist with progress that refreshes while the agent works
What happened, what is risky, what is open?inspectA filterable timeline by lane, severity and session, with decisions
  1. 01

    The agent writes content, never markup. The template, styles and script are fixed, so every page looks the same and passes the same checks.

  2. 02

    The check runs before anything opens: tokens only, focus rings, contrast in both themes, phone width, readable without script, no invented numbers, nothing loaded from the network.

  3. 03

    The script reads the environment and picks where the page goes: an artifact, your browser, a localhost URL with an ssh -L line, or a printed path.

Examples

Real runs, committed in the repo

Ten ComfyUI checkpoints compared on GenEval, weight size, steps and licence, with one row picked
Pick a modelcompare with a pick. One click hands the agent "Use model FLUX.2 [klein] 4B".
A review of three sessions with the open items decided on the page and one reply built for the agent
Review three sessionsinspect: what changed, what is risky, and the open items decided on the page.
The five element planets charted by orbital period, each bar in its element colour, dark theme
Colour that means somethingEach planet wears its element, checked for contrast in light and dark.
Install

Install it in your agent

Any agent that reads Agent Skills
npx skills add IndusCore-AI/intelligent-ui-anything
Claude Code plugin
/plugin marketplace add IndusCore-AI/intelligent-ui-anything
Codex CLI
git clone https://github.com/IndusCore-AI/intelligent-ui-anything ~/.agents/skills/intelligent-ui-anything

Run Codex with --sandbox workspace-write.

What it does not do

  • It does not build or edit your app's UI. It makes throwaway pages that answer one question.
  • No panel inside VS Code yet. Pages open in the browser.
  • Windows, WSL, Cursor and Antigravity paths are written but not yet run.
FAQ

Questions

An open-source Agent Skill that turns what a coding agent produced into one small interactive page (compare, choose, explain, track or inspect) and returns the user's pick to the agent as one line.

Any agent that reads Agent Skills: Claude Code (with or without the Artifact tool), Codex CLI, Cursor, Antigravity and claude.ai. It needs Node.js 18 or later.

Every pick shows the exact line the agent will receive. A Copy button works in every host; when the page is served on localhost, Send posts it straight into the session.

No. It uses Node built-ins only, loads nothing from the network, and writes one self-contained HTML file per page.

"Intelligent UI" is also the name of a ChatGPT feature. This skill is not affiliated with OpenAI. It is MIT licensed and built by IndusCore.

Also from IndusCore: Connect Anything

Open-source agent skills to merge, compile and update Claude Code, Codex and Copilot skills: every source read and scanned, licences checked, losses refused.