Build AI-native apps, not AI demos.

AINative gives you streaming UI, React components, tool execution, and multi-provider support - in one framework.

MIT licensed

TypeScript-first

Zero lock-in

terminal

npm install @hari7261/ainative-client react react-dom

src/main.tsx

import React from "react";
import ReactDOM from "react-dom/client";
import { AIAppComponent, AIPane } from "@hari7261/ainative-client";

function App() {
  const config = {
    apiUrl: "http://localhost:3001",
    streamMethod: "SSE" as const,
  };

  return (
    <AIAppComponent config={config}>
      {(state, app) => (
        <AIPane
          state={state}
          title="Basic Chat"
          subtitle="Powered by AINative"
          onSendMessage={(msg) => app.sendMessage(msg, { stream: true })}
          enableAudio={true}
          enableImage={true}
          enableFile={true}
        />
      )}
    </AIAppComponent>
  );
}

ReactDOM.createRoot(document.getElementById("root")!).render(<App />);

Built for modern teams shipping real AI products

944

npm downloads last 30 days

published npm packages

documented providers

main

active branch

Why AINative

The framework for the next generation of AI apps.

Three primitives that turn LLMs into production software, not chat toys. Each one is independently usable and fully replaceable.

Streaming Native

Token-by-token rendering with backpressure, retry, and resume. Built on the Web Streams API — no polyfills, no opinions.

SSE, fetch, WebSocket
Auto-reconnect
Partial tool calls

Learn more

State-Driven Runtime

A reconciler diffs partial AI output against current UI state and applies minimal patches. No flicker, no duplicate frames, no lost messages.

Optimistic updates
Persistent threads
SSR-safe hydration

Learn more

Provider Agnostic

Start with OpenAI, Anthropic, or Ollama behind one consistent protocol, then extend with your own adapter if you need more.

3 documented providers
Custom adapter API
Per-request routing

Learn more

Components

Composable primitives, not opinionated UIs.

Drop in what you need. Style with your own design system. Every component is headless-compatible and ships with sensible defaults.

AIApp

component

Root provider. Wires runtime, providers, and tools together with a single config object.

<AIApp.Provider app={app}>
  {children}
</AIApp.Provider>

AIInput

component

Multimodal input with attachments, slash commands, and submit-on-enter.

<AIInput
  placeholder="Ask..."
  onSubmit={send}
  accept="image/*,audio/*"
/>

AIStream

component

Renders streaming output with auto-scroll, diffing, and tool-call rendering.

<AIStream
  threadId={id}
  renderTool={Tool}
  autoScroll
/>

AIPane

component

Side-pane assistant for embedded copilots inside any app surface.

<AIPane
  side="right"
  width={420}
  collapsible
/>

Tool Runtime

component

Type-safe tools executed client-side, server-side, or both — with state injection.

app.tool("search", {
  schema: z.object({
    query: z.string(),
  }),
  run: async ({ query }) =>
    searchWeb(query),
})

Multimodal Input

component

Image, audio, and file uploads, normalized into one provider-agnostic format.

<AIInput
  accept="image/*,audio/*"
  maxFileSize={20 * 1024 * 1024}
/>

Examples

One framework. Three runtimes.

The same primitives across React, Node, and Python. Pick the stack you already love.

app/chat.tsx

import { AIApp, AIInput, AIStream } from "@hari7261/ainative-client";

const app = new AIApp({
  provider: "openai",
  model: "gpt-4o",
  endpoint: "/api/ai",
});

export default function Page() {
  return (
    <AIApp.Provider app={app}>
      <AIStream className="flex-1" />
      <AIInput placeholder="Message AINative..." />
    </AIApp.Provider>
  );
}

Architecture

A clean stack, top to bottom.

Every layer is replaceable. Every contract is typed. The same protocol runs from the browser to the model provider.

Layer 01

Browser

Your React app + AINative client runtime

Layer 02

Client Runtime

State manager, event bus, streaming engine, reconciler

Layer 03

Server Adapters

Node Express middleware or Python FastAPI router

Layer 04

Providers

OpenAI, Anthropic, Ollama, plus custom adapters

Layer 05

Tool Registry

Schema-validated functions with state injection

Client⇄Server⇄Providers

One wire protocol. Streamed end to end. Tool calls, state patches, and errors travel through the same typed channel.

Use cases

Ship products that ship value.

From single-purpose chatbots to enterprise copilots — the same primitives scale all the way up.

AI customer support

Tools, citations, and human handoff with full conversation history.

Internal copilots

Embedded assistants that read your data and call your APIs.

Workflow automation

Multi-step agents with retries, branching, and durable execution.

SaaS assistants

Drop-in side panels and inline UI for product surfaces.

Chat interfaces

Beautiful streaming chat with markdown, code blocks, and tools.

Multimodal tools

Image, audio, and document understanding in one input.

Performance

Engineered for production, not slideware.

Every measurement comes from real apps in production — not contrived benchmarks.

<80ms

First streamed token

On warm provider connections from the edge.

Documented providers

OpenAI, Anthropic, and Ollama are the providers listed in the repo README today.

SSR

Server rendering

Hydration-safe streaming primitives, no flicker.

100%

Type-safe API

Zod or JSON schemas in. Inferred types out — everywhere.

12kB

Client gzipped

Tree-shakeable. Pay only for the components you use.

Required services

No vendor cloud, no telemetry, no required accounts.

Loved by builders

Trusted by teams shipping AI to production.

From two-person startups to public companies — AINative is the substrate for serious AI products.

"AINative replaced weeks of glue code. Streaming, tools, and provider switching just work. Our product feels noticeably faster, and our codebase shrank by a third."

Maya Chen

Staff Engineer · Lattice AI

"Streaming UX finally feels native. The reconciler is the part nobody else got right — partial tool calls render perfectly without a single hydration warning."

Diego Alvarez

Founder · Northwind

"We swapped OpenAI for Anthropic in one config line and shipped to prod the same day. Routing requests by user tier was a 4-line change. This is what a real framework looks like."

Priya Raman

Engineering Lead · Helix Labs

Pricing

Open source forever. Pro features coming soon.

The AINative core is MIT-licensed. We're currently building our managed cloud and team tools in private beta.

Free

$0/mo

Open source. Self-host forever.

MIT license
All client packages
All server adapters
Community Discord
Unlimited apps

View Plans

Pro

Enterprise

Custom

SSO, SLA, and dedicated onboarding.

Everything in Pro
Single sign-on (SAML)
Security review
Architecture support
99.9% SLA

Talk to sales

FAQ

Questions, answered.

Don't see yours? Reach out — we read everything.

Ship better AI products with AINative.

Free, open source, and production-ready. Start in under a minute — no signup, no credit card, no telemetry.

Start Building Now Read Documentation

Build AI-native apps, not AI demos.

The framework for the next generation of AI apps.

Streaming Native

State-Driven Runtime

Provider Agnostic

Composable primitives, not opinionated UIs.

AIApp

AIInput

AIStream

AIPane

Tool Runtime

Multimodal Input

One framework. Three runtimes.

A clean stack, top to bottom.

Browser

Client Runtime

Server Adapters

Providers

Tool Registry

Ship products that ship value.

AI customer support

Internal copilots

Workflow automation

SaaS assistants

Chat interfaces

Multimodal tools

Engineered for production, not slideware.

Trusted by teams shipping AI to production.

Open source forever. Pro features coming soon.

Free

Pro

Enterprise

Questions, answered.

Is AINative open source?

Does it support Ollama and local models?

Can I use a Python backend?

Can I self-host AINative?

Does it support streaming?

Can I build production SaaS apps?

How do I migrate from the OpenAI SDK or Vercel AI SDK?

What about observability and logging?

Ship better AI products with AINative.