Vercel's UI SDK gives coding agents building blocks instead of finished components
SummaryUI SDK, from Vercel Labs, is an open-source library of 50+ interface behaviors (popups, typeahead, drag, focus) that agents combine into components for React, Vue, Svelte, Solid or Web Components. It is still a 0.1.0 beta. Not tested by MODEL.
· MODEL Editors · 2 min read
Some links may be affiliate links. Disclosure

What it is. UI SDK is "a library of behaviors for agents to build the next generation of user interfaces," built by Vercel and published as open source (Apache-2.0) at vercel-labs/ui (opens in a new tab). Instead of handing an agent a finished Select, it hands over the pieces a Select is made of: createPopup (open, position, dismiss), createSelection (pick one or many) and createTypeahead (type to jump). The same pieces can then become an autocomplete, a combobox or something new (ui-sdk.dev (opens in a new tab)).
What's in it. The site says it ships 50+ behaviors, among them roving focus, typeahead, popups, selection, reorder, resize, pan-and-zoom viewports and virtualized lists, plus a starter component library (color picker, calendar, OTP input, tree, toolbar and more). Every component works in React, Vue, Svelte, Solid and Web Components. You add the UI SDK skill to your agent, and it picks and combines the behaviors itself. There is also an agent bridge with WebMCP support so agents can read a page and operate its controls directly.
The demo. On Nov. 14, Vercel CEO Guillermo Rauch posted (opens in a new tab) a Winamp-style player built with "two prompts of fx + Gemini Argon" in plain Web Components, which he says reused 14 UI SDK behaviors (window resizing, drag, menubar hover-switching, typeahead, EQ sliders, scoped hotkeys, ARIA, WebMCP agent control). The demo is live at winamp-ui-sdk.vercel.app (opens in a new tab), with source linked.
Who it's for. Front-end teams that let coding agents write UI and want accessible keyboard, focus and screen-reader behavior without the agent reinventing it every time.
The catch. It is early: the @ui-sdk/react package on npm is at 0.1.0 beta (first published Nov. 12, 2026). The claim that agents operating pages this way are "more than twice as fast, at less than half the cost" than with screenshots is Vercel's own, with no independent test yet.
The MODEL take. Giving agents behaviors instead of a fixed component kit is a smart bet on how UI code will get written. Try it on a prototype; wait for a stable release before building a design system on it.
Not tested by MODEL. Not sponsored. No affiliate links.
Get three tested tools every Friday
Free, weekly. Unsubscribe anytime.
Sources
Keep reading

Hot take
Google caught a fully automated AI attack. It didn't need a zero-day. It read the patch.
Google's latest AI Threat Tracker describes exploit_agent, a multi-agent system that diffs web software before and after each patch, drafts an exploit and runs it with no human in the loop. The same report counts hundreds of AI-written exploits that don't work. Our take: sloppy AI code now hits defenders from both sides.
· MODEL Editors

Hugging Face x OpenRouter
Hugging Face now routes its LLM calls through OpenRouter
Hugging Face switched Inference Providers from its own router to OpenRouter's. Same token, same API, a provider picked per request, and retries when one fails. One breaking change: your own provider keys no longer work for chat models. Not tested by MODEL.
· MODEL Editors

Builder tip
Amazon caught AI agents hunting exposed LLM gateways for API keys. Lock yours down.
Amazon's decoy sensors captured attacker AI agents' own campaign records: they scanned for exposed LiteLLM, One API, Dify and similar apps, pulled the provider keys inside, and used stolen keys to pay for more hunting. Every technique was old. Here's the checklist.
· MODEL Editors

xeet
xeet renders X posts to PNG without a browser
Guillermo Rauch's new open-source tool rebuilds X's post layout in Rust and WebAssembly, so apps, agents and CI jobs can preview a post as an image in milliseconds instead of screenshotting headless Chrome. Not tested by MODEL.
· MODEL Editors
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.