hashbrown

Hashbrown

AI chat and agents for your React or Angular app.

Hashbrown is a headless TypeScript framework. The model renders your

components and calls your tools, in the browser, with any provider.

[Angular quick start](https://hashbrown.dev/docs/angular/start/quick) | [React quick start](https://hashbrown.dev/docs/react/start/quick)

hashbrown.dev

Submit an Issue | Contributing Guidelines | Code of Conduct

What is Hashbrown | Installation | How it works | Features | Connect a model | See it in a real app | Contributing

What Is Hashbrown

Hashbrown is a headless TypeScript framework for AI chat and agents in React and Angular. It brings generative UI from your own components, client-side tools, and streaming structured output from any model. Hashbrown ships no chat UI of its own: you keep your components and your design system.

• Generative UI. The model only renders components you register, and Skillet schemas validate their props.
• Tools in the browser. Tools run in your app, with its state and services.
• Built for streaming. Components render while the response streams in.
• Any model. Your server keeps the API key and streams from OpenAI, Anthropic, Gemini, Bedrock, Azure, or Ollama.

Installation

Hashbrown has three kinds of packages:

• @hashbrownai/core: the framework-agnostic primitives, including the Skillet schema language and the streaming JSON parser
• @hashbrownai/: resources, hooks, and components for your framework
• @hashbrownai/: a Node adapter that streams from a model provider's SDK

For Angular and OpenAI:

npm install @hashbrownai/{core,angular,openai}

For React and OpenAI:

npm install @hashbrownai/{core,react,openai}

How It Works

The same API in React and Angular.

1. Expose your components

The model only renders components you register. Skillet validates their props.

Angular:

const components = [
  exposeComponent(FlightCardComponent, {
    name: 'FlightCard',
    description: 'One aircraft on the map, with live altitude and speed.',
    input: {
      note: s.streaming.string('A sentence or two about this flight'),
      hex: s.string('The aircraft hex code from a tool result.'),
    },
    children: false,
  }),
];

React:

const components = [
  exposeComponent(FlightCard, {
    name: 'FlightCard',
    description: 'One aircraft on the map, with live altitude and speed.',
    props: {
      note: s.streaming.string('A sentence or two about this flight'),
      hex: s.string('The aircraft hex code from a tool result.'),
    },
    children: false,
  }),
];

2. Give it tools

Tools run in the browser, with your app's state and services.

Angular:

chat = uiChatResource({
  system: 'Help users understand the planes on the map.',
  components,
  tools: [createTool(this.atc.findAircraft)],
});

React:

export function Assistant() {
  const store = useAtcStore();
  const atc = useMemo(() => createAtcTools({ store, fetchRoute }), [store]);
  const findAircraft = useTool({ ...atc.findAircraft, deps: [atc] });

  const chat = useUiChat({
    system: 'Help users understand the planes on the map.',
    components,
    tools: [findAircraft],
  });

  // Render the stream (step 3).
}

3. Render the stream

Components render while the response streams in.

Angular:

@for (message of chat.value(); track $index) {
  
}

React, at the end of Assistant:

return chat.messages.map((message) =>
  message.role === 'assistant' ? message.ui : message.content,
);

Before you use Hashbrown's hooks or resources, point them at your server.

Angular:

export const appConfig: ApplicationConfig = {
  providers: [provideHashbrown({ baseUrl: '/api/run' })],
};

React:

{children}

Features

Every feature is documented for both frameworks: Angular docs | React docs

Connect a Model

Your server keeps the API key. A Hashbrown adapter maps an AG-UI run to your provider's SDK and streams AG-UI events back to the browser.

• OpenAI: @hashbrownai/openai
• Anthropic: @hashbrownai/anthropic
• Google Gemini: @hashbrownai/google
• Amazon Bedrock: @hashbrownai/bedrock
• Azure OpenAI: @hashbrownai/azure
• Ollama: @hashbrownai/ollama
• Your own backend

Hashbrown's UI packages post to /run by default. You can change the URL as long as the server and the UI use the same one. An Express endpoint with the OpenAI adapter:

import type { RunAgentInput } from '@ag-ui/core';
import { EventEncoder } from '@ag-ui/encoder';
import { HashbrownOpenAI } from '@hashbrownai/openai';

app.post('/run', async (req, res) => {
  const abortController = new AbortController();
  req.once('aborted', () => abortController.abort());
  res.once('close', () => abortController.abort());
  const stream = HashbrownOpenAI.stream.text({
    apiKey: process.env.OPENAI_API_KEY!,
    model: process.env.OPENAI_MODEL ?? 'gpt-5-nano',
    input: req.body as RunAgentInput,
    signal: abortController.signal,
  });
  const encoder = new EventEncoder();

  res.header('Cache-Control', 'no-cache, no-store, must-revalidate');
  res.header('Content-Type', encoder.getContentType());
  res.header('Connection', 'keep-alive');
  res.flushHeaders();

  for await (const event of stream) {
    res.write(encoder.encodeSSE(event));
  }

  if (!res.writableEnded) {
    res.end();
  }
});

Not every model handles every feature. Some older, smaller models struggle with generative UI and tool calling.

See It in a Real App

atc is a live map of airline traffic over the Pacific Northwest with an assistant that answers in your own components: flight cards, arrivals boards and aircraft comparisons. It is built twice, in Angular and React, from one framework-free core. The aircraft data is real, from adsb.lol.

Try the app or run it locally. Put OPENAI_API_KEY=... in the repository's .env, then run each in its own terminal:

nvm use
npm ci
npx nx serve atc-server

npx nx serve atc-react

Open http://127.0.0.1:4342/react/. Use atc-angular and port 4341 for the Angular app. See the example README for the architecture and tests.

Need a Complete Chat UI?

Hashbrown is headless. threadplane is the full agent UI for React and Angular, built on Hashbrown: threads, approvals, and tool progress. Free and MIT, with enterprise support from the team behind Hashbrown.

Core Team

Hashbrown is led by Brian Love. We are always eagerly looking for contributors and collaborators, so if you'd like to help shape Hashbrown, we'd love to hear from you.

Contributing

Hashbrown is a community-driven project. Read our contributing guidelines on how to get involved.

License

MIT © LiveLoveApp, LLC