Skip to content
Response Chat Widget

The chat engine forintelligent commerce

Response Chat is a production-ready widget for brands that want streaming answers, rich product cards, operational record views, and brand-safe escalation paths inside one embeddable chat surface.

OpenAI, Anthropic, and GeminiOrders, returns, subscriptions, and productsCitations, tools, and rich messages
Aria
Streaming live
ChatOrdersProducts
Ask about products, orders, or returns...
provider integrations
3
built-in chat and record views
6
runtime configuration options
100+
drop-in bundle delivery
UMD
Core capabilities

Built for real storefront work

This is not a floating FAQ bubble. Response Chat is a commerce surface with streaming AI, structured interfaces, operational context, and integration hooks that hold up in production.

Streaming replies

Token-by-token SSE responses with optional thinking states, tool calls, tool results, and citations rendered inline.

Commerce cards

Render products, buttons, forms, and cart actions directly in the transcript instead of bouncing customers into a separate flow.

Record views

Built-in tabs for orders, products, subscriptions, returns, and exchanges keep self-service operations inside the widget.

Lifecycle APIs

Thread CRUD helpers, payload builders, response parsers, and imperative open or close controls let you wire the widget into any stack.

Provider flexibility

Run against OpenAI, Anthropic, or Gemini backends with the same route contract and the same front-end surface.

Exports and telemetry

Track message events, widget toggles, outcomes, and conversation exports without forking the component.

Commerce-native messaging

Move from text to action without leaving the thread

Response Chat can switch between free-form answers and structured UI blocks in the same transcript, so customers can browse, verify, and take action without hopping between tabs.

Product cards
Variants, prices, ratings, badges, and add-to-cart actions
Order and subscription cards
Statuses, line items, next actions, and fulfillment details
Buttons and quick replies
Guided flows for shade matching, returns, or escalation
Forms and data capture
Collect order info, support details, or structured intake inside chat
Tool calls and citations
Show work in progress and the sources behind the answer
Markdown and export
Readable long-form answers plus JSON, TXT, or Markdown conversation exports
Operational record view
OrdersReturnsSubscriptions
ORD-1847
3 items · 2-day Express
Fulfilled$127.00
RTN-3008
UPS Ground · RMA active
Pending Inspection$128.00
SUB-4472
Renews Jun 21 · Premium plan
Active$64.00 / mo
embed.html
1<script src="https://unpkg.com/react@18/umd/react.production.min.js"></script>
2<script src="https://unpkg.com/react-dom@18/umd/react-dom.production.min.js"></script>
3<link rel="stylesheet" href="https://stateset.com/chat-widget.css" />
4<script src="https://stateset.com/chat-widget.bundle.js"></script>
5<script>
6 const api = window.ChatWidget.mount({
7 container: document.getElementById('chat'),
8 props: {
9 companyName: 'Your Brand',
10 agentName: 'Aria',
11 messageEndpoint: '/api/chat',
12 threadCreateEndpoint: '/api/threads',
13 threadEndpoint: '/api/threads',
14 enableStreaming: true,
15 enableToolVisualization: true,
16 enableRichMessages: true,
17 enableSourceCitations: true
18 }
19 });
20</script>

Integration surface

Mount it once. Wire it to your stack.

Response Chat ships as a UMD bundle with a small imperative API, sample servers, and request builders that let you reshape thread and message traffic without forking the UI.

  • Works in plain HTML, React, Vue, Shopify themes, and Next.js
  • Supports thread create, update, delete, and fetch flows
  • Remote settings and custom CSS can tailor the widget per brand
  • Callbacks expose send, receive, error, thread, and outcome events

Production guardrails

Ready for brand, security, and support teams

Response Chat was built as a product surface, not a prototype. The widget includes the controls you need before putting AI in front of real customers.

Widget auth tokens

Sign instances with widget tokens before they can talk to your backend.

Domain allowlisting

Restrict where the widget can be embedded across brand and partner properties.

CSP and consent

Nonce-aware scripts and consent gates keep the widget aligned with stricter policies.

Accessible by default

Keyboard support, ARIA labels, focus handling, and live-region semantics are built in.

Global rollout

Locale packs, translation overrides, and automatic RTL handling are available at runtime.

Analytics hooks

Track widget_loaded, chat_toggle, message_sent, outcomes, and feedback events.

Thread persistence

Persist sessions and thread IDs so customers can resume without starting over.

Live agent handoff

Escalate to a human while preserving context and keeping the transcript intact.

Where it fits

A better front door for commerce conversations

The widget can sell, support, and escalate in the same surface. That makes it useful for both growth teams and service teams instead of forcing separate chat products for each.

Storefront concierge

  • Guide shoppers to the right product or variant
  • Render product cards with pricing, badges, and actions
  • Keep FAQs and recommendations inside the same thread
  • Use citations when the answer comes from your knowledge base

Post-purchase assistant

  • Surface order status, returns, exchanges, and subscriptions
  • Show operational next steps without sending shoppers to email
  • Collect structured input for support and self-service workflows
  • Resume the thread with persistent context on the next visit

Brand-trained advisor

  • Tune tone, welcome copy, quick replies, and escalation rules
  • Theme the widget per brand with runtime config and CSS
  • Deploy specialized flows for beauty, apparel, wellness, or support
  • Hand off to a live agent when AI should stop and a human should step in

FAQ

Questions teams ask before rollout

The implementation surface is flexible, but the operating model is straightforward.

Load the stylesheet and UMD bundle, then call window.ChatWidget.mount() with your endpoints and props. The same widget can sit inside a storefront, help center, or account portal.

Ship chat that can actually close the loop

Give shoppers answers, actions, and escalation paths in one place instead of sending them through email, ticket forms, and dead-end chatbots.