drop in an agent.
ship it.
the only batteries-included agent component. tools, streaming, approvals, widgets — all wired to a runtime that handles the hard parts. not a chat template. a complete solution.
npx shadcn@latest add https://ui.inference.sh/r/agent.jsontry it
an agent with client-side tools. it can read and fill the form. no backend required for the tools.
how can I help?
I can help configure your pricing plan
examples
agents return structured output. the widget renderer turns it into interactive ui.
departure
SFO
san francisco5h 45m
on timearrival
JFK
new yorksan francisco
partly cloudy
68°
H: 72° L: 58°
showcase
rich interfaces from json. no custom components needed.
AAPL
$178.72
apple inc. • nasdaq
the weeknd
focus time
24:59
payment sent
$250.00
components
agent-native ui. tools, approvals, widgets — built in, not bolted on.
agent
one component. runtime included. tools, streaming, approvals, widgets.
chat ui
building blocks: container, messages, input, and indicators.
markdown
rich markdown with syntax highlighting and embeds.
widgets
declarative ui from json - forms, buttons, cards.
tool ui
tool lifecycle: pending, progress, approval, results.
task output
real-time task display with streaming support.
quick start
one component. four steps. production-ready agent ui.
install the agent component
1npx shadcn@latest add https://ui.inference.sh/r/agent.jsonadd the api proxy route
1// app/api/inference/proxy/route.ts2import { route } from '@inferencesh/sdk/proxy/nextjs';3export const { GET, POST, PUT } = route;set your api key
1INFERENCE_API_KEY=inf_...use the component
1<Agent2 proxyUrl="/api/inference/proxy"3 agentConfig={{4 core_app: { ref: 'openrouter/claude-haiku-45@1ps10tmc' },5 description: 'a helpful ai assistant',6 system_prompt: 'you are helpful.',7 }}8/>