BigCommerce x Cloudflare Hackathon

Build agentic commerce apps on the Developer Platform

A 90-minute kickoff for teams building AI shopping agents, code-generating storefronts, multimodal returns flows, and durable commerce workflows.

Workers Agents SDK Workers AI Sandbox SDK Live BigCommerce path

90-minute flow

Agenda

0:00-0:08

Why Cloudflare plus BigCommerce

North star, live-capable services, and what teams should ship by the end.

0:08-0:18

Platform map

Workers, Agents, AI, Sandbox, R2, Images, Workflows, Browser Run.

0:18-0:50

Tier 1 demos

Photo Studio, Storefront Builder, Voice Shopping, Returns Agent.

0:50-1:05

Tier 2 idea bank

Concierge, Data Analyst, Catalog Enricher, A/B Test Generator.

1:05-1:20

Breakout build plan

Pick a pattern, run locally, swap fixtures only after the UI works.

1:20-1:30

Questions and launch

Hub link, repo map, mentors, and next steps.

How to think about the kit

Live-capable commerce, fallback-ready demos

Worker UI
Workers AI
Agents and Durable Objects
R2, KV, D1
BigCommerce APIs

Start live-capable

Each app names the Worker endpoint, Workers AI call, and BigCommerce seam that can run with credentials.

Keep fallbacks

Fixture data and review logs stay available only for missing credentials, quotas, latency, or conference Wi-Fi risk.

Ship a slice

Teams should prove one durable commerce moment before expanding the surface area.

Tier 1 demo 01

Product Photo Studio

Generate campaign-ready product photos through a Workers AI runtime endpoint, store variants in R2, serve optimized crops through Images, and publish the selected asset back to the catalog path.

POST /api/generate-imageWorkers AIR2Catalog REST

Hero flow

  1. Select a product.
  2. Enter a lifestyle prompt.
  3. POST to /api/generate-image.
  4. Store selected output in R2.
  5. Serve transformed crops through Images.
  6. Publish or fallback-log the catalog update.

Architecture

Photo Studio service path

flowchart LR
  Merchant[Merchant browser] -- prompt --> Worker[Cloudflare Worker SSR UI]
  Worker --> Storefront[BigCommerce Storefront GraphQL]
  Fixtures[Fixture product fallback] -. fallback .-> Worker
  Worker --> AI[Workers AI image model]
  AI --> R2[R2 generated asset bucket]
  R2 --> Images[Cloudflare Images transforms]
  Images --> Worker
  Worker --> Catalog[BigCommerce Catalog REST API]
  Catalog -. fallback .-> Log[Reviewable activity log]

Tier 1 demo 02

Storefront Builder

A merchant describes a campaign and gets a working, previewable storefront page from a Cloudflare-native Sandbox Storefront Builder backed by BigCommerce catalog context.

POST /api/generateSandbox SDKWorkers AIStorefront GraphQL

Hero flow

  1. Prompt a campaign page.
  2. Inject Storefront GraphQL context.
  3. Generate route and components with Workers AI.
  4. Run preview in Sandbox.
  5. Iterate from merchant feedback.
  6. Deploy or export.

Architecture

Storefront Builder generation loop

flowchart LR
  Merchant[Merchant prompt] --> Worker[Worker-hosted studio UI]
  Worker --> Context[BigCommerce catalog context]
  Context --> Generator[Workers AI generation endpoint]
  Generator --> Files[Generated routes and components]
  Files --> Sandbox[Cloudflare Sandbox preview]
  Sandbox --> Deploy[Worker or WFP deploy]
  Fixtures[Fixture product fallback] -. fallback .-> Context
  Worker --> Generator

Tier 1 demo 03

Voice Shopping Agent

A persistent per-shopper Durable Object that supports multilingual voice shopping, product recommendations, spoken responses, and cart updates.

Agents SDKDurable ObjectsRealtimeKitNova-3 STTAura-1 TTS

Hero flow

  1. Speak a shopping need.
  2. Show transcript.
  3. Recommend products.
  4. Speak the response.
  5. Add the best item to cart.
  6. Refresh to prove persistence.

Architecture

Voice shopping session path

flowchart LR
  Shopper[Shopper browser] --> UI[Worker-served phone UI]
  UI --> RT[RealtimeKit WebRTC session]
  RT --> Agent[AIChatAgent Durable Object]
  Agent --> STT[Workers AI Nova-3 STT]
  Agent --> Tools[BigCommerce tool adapter]
  Tools --> BCTools[bc-tools package]
  BCTools --> Storefront[BigCommerce Storefront API]
  Agent --> TTS[Workers AI Aura-1 TTS]
  TTS --> RT
  Agent --> State[(Durable Object shopper state)]

Tier 1 demo 04

Returns Agent

A multimodal returns agent classifies product damage through a Workers AI vision endpoint, decides whether approval is needed, processes live or fallback refunds, and schedules follow-up work.

POST /api/returns/classifyWorkers AI VisionR2Human approvalthis.schedule()

Hero flow

  1. Upload damage photo.
  2. POST to /api/returns/classify.
  3. Offer refund or replacement.
  4. Auto-approve low value.
  5. Escalate high value.
  6. Schedule a 7-day follow-up.

Architecture

Returns decision and follow-up path

flowchart LR
  Customer[Customer browser] --> Worker[Returns Agent Worker]
  Worker --> Upload[Upload/dropzone UI]
  Upload --> API[POST /api/returns/classify]
  API --> Vision[Workers AI vision]
  Vision --> Decision[Policy and return decisioning]
  API --> R2[(R2 return evidence)]
  Decision --> HITL{High-value refund?}
  HITL -->|Yes| Approval[Human approval queue]
  HITL -->|No| BigCommerce[BigCommerce Admin API]
  BigCommerce -. fallback .-> Log[Reviewable refund log]
  Approval --> BigCommerce
  Decision --> Schedule[this.schedule follow-ups]
  Schedule --> Agent[Future AIChatAgent]

Tier 2 idea bank

Four compact demos teams can remix

Shopping Concierge

Text chat shopping agent with product recommendations, preference state, cart previews, and BigCommerce tool telemetry.

AIChatAgentDO/KV/D1

Data Analyst

Merchant asks natural language questions and Sandbox runs Python to create KPIs, charts, and artifacts.

SandboxR2

Catalog Enricher

Bulk SEO titles, meta descriptions, alt text, tags, QA status, and durable publish workflow through a Workers AI enrichment endpoint.

POST /api/catalog/enrichWorkers AIWorkflows

A/B Test Generator

Generate landing page variants, capture Browser Run screenshots, and score the review board.

Browser RunWorkers AI

Tier 2 architecture patterns

Compact service maps

flowchart LR
  Shopper --> Chat[Concierge /api/chat]
  Chat --> Agent[AIChatAgent]
  Agent --> Tools[bc-tools]
  Agent --> State[(DO/KV/D1)]
flowchart LR
  Merchant --> Analyst[Data Analyst Worker]
  Analyst --> Sandbox[Python Sandbox]
  Sandbox --> R2[(R2 artifacts)]
  R2 --> Dashboard[Charts and tables]
flowchart LR
  Catalog --> Agent[Enrichment Agent]
  Agent --> Workflow[Workflow per SKU]
  Workflow --> AI[Workers AI]
  Workflow --> Admin[Catalog REST]
flowchart LR
  Brief --> AI[Variant generation]
  AI --> Sandbox[Sandbox preview]
  Sandbox --> Browser[Browser Run]
  Browser --> R2[(Screenshots)]

Build guidance

Recommended breakout path

1

Run local

Install one repo, run Wrangler, and verify the live endpoint shape before changing code.

2

Pick one seam

Wire exactly one live seam: AI call, storage binding, tool adapter, workflow, or BigCommerce credential path.

3

Demo the moment

Show the commerce outcome with fallback data ready if live services are slow.

Launch resources

Scan the hub, pick a repo, start building