The Infrastructure for Agentic Commerce Just Arrived

What UCP Actually Is

UCP defines how AI agents interact with merchants to browse, checkout, pay, and manage orders - without requiring custom integrations for every connection.

The protocol includes:

• Native MCP binding - the same protocol Claude and other AI assistants already use
• Agent-to-Agent (A2A) support - so agents can coordinate with each other
• OAuth 2.0 identity linking - allowing agents to maintain secure, authorized relationships with merchants
• AP2 payment integration - cryptographically verified consent for transactions

In plain terms: this is the plumbing that lets an AI assistant buy something on your behalf, from any participating merchant, using your preferred payment method - securely and at scale.

The Divergence Is Real

We've been operating on a thesis we call The Divergence - the idea that commerce is splitting into two distinct channels: agent-mediated and human-mediated.

Not replacing. Splitting.

Some purchases will remain deeply human experiences. You'll still want to walk through a showroom, touch materials, talk to someone who understands lighting design. That's not going away.

But other purchases - reorders, replenishment, researched commodities, spec-driven procurement - increasingly won't require a human in the loop. An agent with the right context, permissions, and access can handle them faster and more accurately than a person toggling between browser tabs.

UCP is the strongest validation we've seen that the major players agree with this thesis. They're not waiting to see if it happens. They're building for it now.

The Protocol Stack: How These Pieces Fit Together

UCP doesn't operate in isolation. It's part of an emerging stack of agentic protocols, each handling a different layer:

The power isn't in any single protocol - it's in the combination.

An interior designer's AI assistant uses MCP to access her project specifications and client preferences. It queries a showroom's product catalog (also via MCP). When she approves a fixture selection, the assistant initiates checkout via UCP, processes payment through AP2, and the showroom's inventory agent confirms fulfillment via A2A.

Four protocols. One seamless transaction. No phone calls, no PDF quotes emailed back and forth, no manual PO entry.

This is where commerce is heading. The only question is which platforms will be ready.

What This Looks Like for Our Clients

For Lighting Showrooms

The showroom experience isn't going away - designers and homeowners will still want to see fixtures in person, consult with experts, and make aesthetic decisions with human guidance.

But the back-end of the business? That's ripe for agent-mediated efficiency.

Scenario: The Repeat Builder

A general contractor works with your showroom on every project. Today, their office manager calls or emails to reorder the same recessed lighting package they've spec'd on the last twelve builds. Your team manually enters the PO, confirms inventory, sends a quote, waits for approval, processes payment.

Tomorrow: The GC's procurement agent already knows their standard specs, preferred pricing tier, and payment terms. It queries your catalog, confirms availability, places the order, and handles payment - all before your team's first cup of coffee. Your staff focuses on the complex projects that actually need human expertise.

Scenario: The Designer's Assistant

An interior designer is specifying fixtures for a renovation. Her AI assistant has access to the project's mood board, room dimensions, and budget parameters. It searches your catalog for options that match, filters by availability and lead time, and presents her with a shortlist. She makes the creative decision. Her agent handles the transaction.

You just made a sale without a single email thread.

For Furniture Retailers

The dynamics are nearly identical. High-touch showroom experiences for customers making aesthetic decisions. Agent-mediated efficiency for the transactional layer.

Scenario: The Home Stager

A staging company furnishes dozens of properties per month. Their AI knows the inventory they typically use, preferred vendors, and budget per square foot. When a new property comes in, the agent pulls together a furniture package from your catalog, checks availability across warehouses, and places the order - routed for human approval only if something's out of stock or over budget.

Scenario: The Trade Program

Your trade customers - designers, architects, developers - already expect streamlined ordering. Agent-mediated commerce takes "streamlined" to a different level. Their AI maintains an authorized relationship with your platform (via OAuth), knows their discount tier, and can transact on their behalf within parameters they've set. You get more trade orders with less friction. They get faster fulfillment with less admin work.

For Cannabis Operators

Cannabis retail has unique constraints - compliance, inventory tracking, customer verification - but the core dynamic still applies. Some purchases benefit from human interaction. Others don't.

Scenario: Compliant Reordering

A dispensary's inventory agent monitors stock levels, knows the regulatory requirements for the state, and maintains authorized relationships with licensed distributors. When flower inventory drops below threshold, the agent initiates a compliant reorder - proper documentation, verified licensing, tracked chain of custody - without a buyer spending two hours on phone calls and paperwork.

Scenario: The Informed Consumer

A customer's AI assistant knows their preferences, tolerance levels, and what they've purchased before. It queries your menu, filters for products matching their profile, checks current availability, and presents options. The customer walks in knowing exactly what they want - or places an order for pickup without browsing at all.

What This Means for B2B and Specialty Retail

Here's what's interesting: look at the UCP endorsement list. It's dominated by B2C giants and payment processors.

B2B verticals - lighting, furniture, building materials, specialty wholesale - aren't represented yet. But that doesn't mean they're exempt from this shift. It means they're behind.

Most platforms serving these industries run on legacy architecture. They weren't built for API-first commerce, let alone agent-mediated transactions. When their customers start expecting UCP compatibility (and they will), these platforms will face a choice: rebuild from scratch or become irrelevant.

The showrooms, retailers, and operators we work with may not be thinking about agentic commerce today. But the contractors, designers, procurement teams, and consumers they sell to? They're already using AI assistants daily. The gap between "I use ChatGPT for research" and "my AI handles reorders" is smaller than most realize.

Where We Go From Here

We're not announcing a UCP integration today. The protocol is new, and the spec will evolve.

But we've been building with this future in mind:

• API-native architecture that can expose commerce capabilities to agents via Neural Core
• MCP tooling already integrated into our platform
• AI-enriched product data structured for machine readability as much as human browsing
• Normalized catalog feeds designed to be queried programmatically

When lighting showrooms, furniture retailers, and cannabis operators need to participate in agent-mediated commerce, we intend to be the platform that makes it possible - without requiring them to rebuild everything they've already built with us.

The infrastructure just arrived. The question for every commerce platform is: were you already building toward this, or do you need to start over?