Autonomous Multi-Agent Commerce with Verifiable Digital Credentials
Built by Diego Zuluaga — Solution Architect at Futurewei Technologies (Open Mobile Hub / Linux Foundation)
Nebius.Build SF Hackathon 2026 | Problem Statement #1: Edge Inference & Agents
Every online purchase requires a human in the loop — browsing, comparing prices, checking if a seller is legit, negotiating. AgentCommerce eliminates that entirely.
Two AI agents — a Buyer and a Merchant — negotiate purchases end-to-end with zero human intervention. The Buyer researches real-time market prices, verifies the merchant's identity through digital credentials, and negotiates the best deal below market price. The Merchant responds with counter-offers, price-matches competitors, and closes the sale — all in under 60 seconds.
| Feature | Description |
|---|---|
| Heterogeneous Multi-Model Negotiation | Buyer runs on Nebius AI Studio (MiniMax-M2.1), Merchant runs on OpenRouter (MiniMax-M2). Two different providers, one seamless transaction. |
| Verifiable Digital Identity | Both agents exchange and verify ISO 18013-5 mdoc credentials — the same standard behind mobile driver's licenses — before any transaction. |
| Real-Time Price Intelligence | Tavily Search pulls live pricing from nike.com, garmin.com, rei.com, and other retailers. The buyer never pays above market price. |
| Live Visualization | Every step streams via WebSocket to a React dashboard showing the full negotiation arc in real-time. |
REACT FRONTEND
(Live Dashboard via WebSocket)
|
WebSocket
|
FASTAPI BACKEND
Orchestration Engine
/ \
BUYER AGENT MERCHANT AGENT
Nebius AI Studio OpenRouter
MiniMax-M2.1 MiniMax-M2
| |
Tavily Search ISO 18013-5 mdoc
(Market Price Data) (Credential Verification)
DISCOVERY → RESEARCH → CREDENTIAL_EXCHANGE → NEGOTIATION → AGREEMENT → PAYMENT → CONFIRMED
Each state transition emits a WebSocket event for real-time visualization.
Example Transaction:
- Garmin Forerunner 265 GPS Watch listed at $450
- Tavily verifies market price at $349.99
- Agents negotiate over 3 rounds
- Deal closed at $315 — 10% below market price
- Both identities verified via ISO 18013-5 mdoc credentials
| Layer | Technology | Purpose |
|---|---|---|
| Buyer Agent LLM | Nebius AI Studio (MiniMax-M2.1) | Hackathon host sponsor |
| Merchant Agent LLM | OpenRouter (MiniMax-M2) | Multi-provider interop |
| Market Research | Tavily Search API | Real-time price verification |
| Identity | ISO 18013-5 mdoc / DPC mock engine | Digital credential exchange |
| Backend | Python 3.11+ / FastAPI / WebSockets | Async orchestration |
| Frontend | React + Vite + TailwindCSS | Live dashboard |
| Package Manager | uv | Python dependency management |
- Python 3.11+
- Node.js 18+
- uv package manager
# Clone
git clone https://github.com/dzuluaga/agentcommerce.git
cd agentcommerce
# Environment variables
cp .env.example .env
# Edit .env with your API keys:
# NEBIUS_API_KEY=your_nebius_ai_studio_key
# OPENROUTER_API_KEY=your_openrouter_key
# TAVILY_API_KEY=your_tavily_key
# Backend
uv sync
uv run uvicorn backend.main:app --reload --port 8000
# Frontend (new terminal)
cd frontend
npm install
npm run devOpen http://localhost:5173 and click a preset to start a transaction.
- Discovery — Buyer agent identifies products from merchant inventory matching the purchase goal
- Research — Tavily searches official retailer sites for verified market prices (MSRP)
- Credential Exchange — Both agents present and verify ISO 18013-5 mdoc digital credentials
- Negotiation — Up to 5 rounds of offers and counter-offers, informed by real market data
- Agreement — Agents converge on a price below market value
- Payment — Buyer presents payment credential
- Confirmed — Transaction complete with full audit trail
- Nebius AI Studio — Powers the Buyer agent's LLM inference (MiniMax-M2.1)
- OpenRouter — Powers the Merchant agent's LLM inference (MiniMax-M2)
- Tavily Search — Provides real-time market price data from official retailer websites
MIT
Built for the Nebius.Build SF Hackathon — March 15, 2026