On January 11, 2026, Google announced the Universal Commerce Protocol (UCP) at the National Retail Federation conference.

UCP is an open standard designed to solve the N×N integration problem in agentic commerce.

Right now, every AI platform that wants to enable shopping needs custom integrations with every merchant. ChatGPT built one for Shopify, Gemini is working on another and Microsoft Copilot is working on integrations too.

Each merchant implements the same checkout logic three different ways.

UCP collapses this complexity by being the one integration applicable across all AI platforms.

The protocol handles the full commerce journey: product discovery, cart management, checkout, discount application, loyalty programs, and post-purchase support.

I built a UCP-compliant record store to see how Google designed for these constraints.

1. mactop - Apple Silicon Monitor Top

Terminal-based monitoring tool for M-series chips with real-time CPU/GPU usage, power consumption, temperatures, and Thunderbolt bandwidth. No sudo required as it uses native Apple APIs with 15 layouts, process management, and JSON export for scripting.

2. npmgraph - Visualize NPM Dependencies

Interactive dependency graphs for any NPM package with collapsible subtrees and colorization options. Useful for auditing bloated dependencies and identifying circular deps before adding packages.

3. BentoML - Production ML Model Serving

Open-source framework for serving AI/ML models in production with auto-scaling, batching, and adaptive optimisation. Package any model (PyTorch, TensorFlow, vLLM) into a standardised format and deploy anywhere.

Let’s say a merchant publishes a capability file at a well-known URL on their domain. Any AI agent hits that URL and learns what the store supports: checkout, discounts, shipping. The agent then transacts using standard endpoints.

Checkout is three API calls: create a cart, optionally apply discounts, complete the purchase. The same three calls work for any UCP-compliant store.

This is a deliberate architectural decision. Instead of agents learning custom APIs for each merchant, they query a standard capability file and adapt to what each merchant offers.

A furniture retailer might require delivery date selection, but a subscription service might offer billing cadences. The protocol accommodates both without breaking.

Step 1: set up your product catalog

A UCP merchant needs products to sell and, optionally, discount codes. These live in CSV files that get loaded into a SQLite database.

Prices are in cents; $35.00 becomes 3500. The stock is tracked per item, discount codes have types (percentage or fixed) and expiry dates the server enforces.

Step 2: install and start the server

UCP provides a Python SDK and a ready-to-run FastAPI server.

Clone both, load your catalog, and start it:

Your store is live on localhost:8182 with five endpoints.

Step 3: capability discovery

An AI agent hits the store's well-known URL to learn what it supports:

Same pattern as .well-known/openid-configuration in OAuth. The agent now knows this store supports checkout and discount codes and accepts Shop Pay.

Notice the extends field: discount builds on top of checkout.

Capabilities compose instead of being a flat list.

Step 4: create a checkout session

Add a record to the cart:

The agent sends both the item ID and the price, the server cross-checks both against the database before creating the session and status starts at incomplete.

The server validates the price. Try sending "price": 1000 for a record that costs $35, the server will reject it.

An AI agent can hallucinate a cheaper price; UCP ensures the merchant is always the price authority.

Step 5: apply a discount code

10% off $35.00 = $3.50 discount. New total: $31.50.

The allocations array uses JSONPath to map every discount dollar to a specific line item. In a bigger cart with a "10% off jazz only" code, the agent would know exactly which records were discounted and by how much.

The server also checks expiry dates. Send an expired code, and the applied array comes back empty. There are no silent failures and the agent knows exactly what happened and can tell the user.

Step 6: complete the purchase

incomplete → completed is irreversible.

Try completing the same session again; the server rejects it. This prevents duplicate charges, which is trivial with stateless tool calls but critical when money is moving.

In production, the agent needs to handle failures at each transition: expired discount at step 5, declined card here, out-of-stock between steps 4 and 6.

Your record store is UCP-compliant. Any AI agent that speaks UCP can now discover it, browse it, and buy from it, using the same protocol Shopify's million+ merchants, Target, and Walmart use.

UCP's protocol design is solid.

The open question now for me is its adoption outside Google's ecosystem.

OpenAI built their own commerce protocol with Stripe and Microsoft has Copilot Checkout.

Economics favours convergence so no merchant wants to implement three different shopping protocols, but the winner isn't decided yet.

Shopify co-developing UCP is significant. They process billions in transactions and serve millions of merchants so they have the distribution.

If UCP becomes Shopify's standard interface, every Shopify merchant becomes UCP-compatible by default.

So it’s worth building with right now.

And even if the protocol wars continue, the primitives are still right.

Until next time,
Vaibhav 🤝🏻

If you read till here, these might be interesting:

#AD 1

AI gets smarter when your input is complete. Wispr Flow helps you think out loud and capture full context by voice, then turns that speech into a clean, structured prompt you can paste into ChatGPT, Claude, or any assistant. No more chopping up thoughts into typed paragraphs. Preserve constraints, examples, edge cases, and tone by speaking them once. The result is faster iteration, more precise outputs, and less time re-prompting. Try Wispr Flow for AI or see a 30-second demo.

Start flowing free

#AD 2

Most read the move after it runs. The top 250K start before the bell.

Elite Trade Club turns noise into a five-minute plan—what’s moving, why it matters, and the stocks to watch now. Miss it and you chase.

Catch it and you decide.

Get the Next Alert

_{By joining, you’ll receive Elite Trade Club emails and select partner insights. See Privacy Policy.}