Editorial / Thesis-led / Independent
The commerce layer
AI agents are going to read.
Online commerce is being rewritten for a new kind of client — autonomous shoppers, LLM assistants and agent-pay rails. This is the reference on the emerging infrastructure that makes catalogs, offers and policies machine-readable, interoperable and transactable. Call it a Universal Commerce Protocol.
- Focus
- Commerce infrastructure for AI agents — not generic SEO, not `llms.txt`, not vendor marketing.
- Posture
- Platform-neutral, operator-empathetic, maturity-aware (Established / Emerging / Speculative).
- Readers
- E-commerce operators, platform PMs, catalog engineers, analysts and agent builders.
The thesis
A new protocol layer is forming between online commerce and AI systems.
- 01
Humans are no longer the only clients of commerce data.
For two decades, product data optimized for shoppers and search. Agents are a third, structurally different consumer — reading at machine speed, reasoning about intent, acting on behalf of the buyer.
- 02
The commerce stack is missing a protocol layer.
There is no shared contract between merchants, agents and platforms for offer, stock, policy and intent. Composite emergence is underway — payment rails, agent tooling, catalog semantics, identity.
- 03
Merchant readiness is the decisive variable.
Platforms will ship most of the plumbing. Catalog quality, attribute richness and policy clarity remain operator responsibilities — and that is the wedge.
- 04
Discovery economics change when the intermediary is an agent.
Being the answer, not the link. Semantic addressability is the new surface area. SEO becomes a precondition, not the finish line.
Pillars
Six angles on the agentic commerce stack.
Definition, problem, actors, systems, data and map. Each pillar links down into operator-grade satellites and up into the thesis.
Definition
What is the Universal Commerce Protocol
The conceptual frame for the infrastructure layer that makes catalogs, offers and policies readable and actionable by AI agents.
Read →
Problem
Why commerce needs machine-readable infrastructure
Human-optimized product pages and SEO are insufficient when agents are the new clients of commerce data.
Read →
Actors
AI agents and the future of e-commerce
How autonomous shoppers, LLM assistants and agent-pay rails change discovery, checkout and after-sales.
Read →
Systems
Commerce interoperability, explained
The shared semantics and patterns required for heterogeneous catalogs to reach heterogeneous agents.
Read →
Data
Product catalogs for AI systems
Turning a PIM feed into an agent-retrievable, policy-aware, semantically addressable product surface.
Read →
Map
Standards, schemas and protocols
schema.org, GS1, MCP, A2A, Apps SDK, Stripe ACP, Visa IC — the emerging stack, compared.
Read →
Map
No single spec. A composite stack.
"Universal Commerce Protocol" is the conceptual convergence of payment rails, agent tooling, catalog semantics, identity and trust. We map and name it — openly and neutrally.
See the full standards comparison →- Schema.org Product / Offer Catalog semantics on the open web.
- GS1 / GTIN Global trade item identity.
- Google Merchant Center De facto feed standard for commerce discovery.
- MCP (Model Context Protocol) Agent-to-tool context exchange.
- A2A (Agent-to-Agent) Horizontal agent cooperation, emerging.
- Stripe Agentic Commerce Agent-ready payment intents and tokens.
- Visa Intelligent Commerce Agent-bound card credentials.
- OpenAI Apps SDK Surface applications inside the ChatGPT client.
Operators
Is your catalog agent-ready?
A 47-point readiness checklist across catalog semantics, policies, feeds, identifiers, agent-discovery, and transactional hygiene. Work through it with your PIM, feed and merchandising teams.
- ✓ Product identifiers (GTIN, MPN, brand)
- ✓ Offer semantics (price, availability, shipping windows)
- ✓ Policies as structured data (returns, warranty)
- ✓ Attributes beyond marketing copy
- ✓ Discoverability signals for agents
- ✓ Agent-pay readiness (tokens, scopes)
- ✓ After-sales traceability
Resources
Long-form essays and working notes.
From SEO to agent visibility: a mental model
Why ranking for humans and being retrieved by agents are related but distinct practices.
Read →
Anatomy of an agent purchase
A step-by-step decomposition of a shopping agent purchase, from intent to confirmation.
Read →
Catalog as API
Treating the product catalog as a deliberate API surface — versioned, typed, documented.
Read →
We publish what operators will need to read before the next platform release notes drop.
Browse all resources →