Headless commerce · Vercel

Your composable storefront, now with an agentic layer that converts.

AI shopping agents don’t just list products — they discover, compare, and recommend. To a headless storefront they see a rendered SPA with no machine catalog and no decision signals, so they can’t recommend you.

6
AI shopping platforms
5
commerce backends in
15
decision-intelligence fields / product
<15 ms
p99 edge reads
The headless gap

Agents can’t recommend what they can’t read.

Headless gives your team the frontend you want — your framework, your design system, your edge. But the catalog lives behind an API, and the page an agent fetches is rendered HTML. There’s no machine surface to query and no decision signals to read, so the agent scrapes one product at a time — or moves on to a store it can.

Even the base ACP/UCP protocol only hands an agent a flat product list. To be recommended, you have to be discoverable, deeply explorable, and decision-ready — answering the questions an agent asks before it picks.

Inception adds that layer at the edge you already deploy to. The frontend stays yours.

What an AI agent gets from your store
Stock headless
  • Rendered SPA, no catalog
  • Scrape one product at a time
  • No structured comparisons
  • No trade-offs or fit signals
  • No suggested queries
  • Invisible to recommendations
With Inception Agents
  • Your normalized catalog
  • One query, whole catalog
  • Structured comparisons
  • Trade-offs + use-case fit
  • Suggested query expansion
  • Recommended with confidence
What you get

Four layers that turn a headless SPA into a recommendation.

One install wires all four. They’re built to work together — your catalog, the intelligence to recommend it, the attribution that proves it, and the discovery that surfaces it.

CATALOG

Catalog route-through

Your real, normalized catalog served from your own domain when an agent asks for it — not a static brochure.

INTELLIGENCE

Decision intelligence

Comparison narratives, trade-offs, use-case fit, and ideal-buyer profiles on every product — the signals agents use to choose.

TRACE

Trace + intent graph

Every agent-driven arrival stitched and compounding into a per-merchant intent graph that sharpens over time.

DISCOVERY

Edge discovery

llms.txt, JSON-LD, and your /agent/* endpoints served sub-15ms at your edge, so agents find and read you first.

The catalog route-through

Your dynamic agent surfaces, served from your real catalog.

Discovery files tell an agent you exist. The route-through lets it explore. When an agent hits a dynamic surface, the middleware signs the request and forwards it to our engine, which resolves your store by hostname and serves your normalized catalog — at your own domain.

01
Request
An agent hits a dynamic surface
A shopping agent requests /agent/query, /agent/products, or your ACP/UCP endpoints on your own domain — the surfaces that need your live catalog.
02
Sign
The middleware signs and forwards
withInception HMAC-SHA256-signs the request with your per-tenant key and forwards it — path and query intact — to the Inception engine.
03
Verify
The engine verifies and resolves you
It verifies the signature, binds the request to your tenant by host (no cross-tenant access), and looks up your normalized catalog.
04
Serve
Your catalog + intelligence are served
The agent receives your real products enriched with the decision intelligence to recommend them. On any failure, the request degrades to your origin — never blocked.
Any backend in

Your catalog comes from the backend you already run — normalized into one canonical product schema. Vercel is the delivery layer, not the catalog source. N backends × M agent platforms collapse to N + M through a single normalization layer.

Shopify
OAuth · webhooks
WooCommerce
REST API · HMAC
BigCommerce
Multi-storefront
Adobe Commerce
Magento 2 · I/O Events
Salesforce Commerce Cloud
SFCC · enterprise

Signed with your per-tenant key · host-bound to your tenant · fails safe to your origin.

Beyond the protocol

ACP and UCP are table stakes. Recommendations are won on intelligence.

Any feed can list products. Inception enriches every product with the decision-support layer agents actually use to choose — and exposes the query surfaces that turn a glance into deep exploration. The protocol gets you in the room; this is how you win it.

COMPARISON NARRATIVES

Why you, over them

Every product carries structured why-choose-over claims, the alternatives it competes with, and a decision narrative — so an agent weighing options has your side of the story, in its own words.

TRADE-OFFS & LIMITS

Honest by design

Key trade-offs, honest limitations, and who a product is not ideal for. Agents reward honest data — stated limitations read as a trust signal and earn more recommendations than optimistic copy.

FIT & IDEAL BUYER

Scored for the shopper

Per-use-case fit scoring, best-for scenarios, and an ideal-buyer profile — so an agent carrying a shopper’s context can match the right product to the right person.

PURCHASE GUIDANCE

Decision support, not specs

Purchase guidance and value rationale that help an agent reason about the choice — the nuance a great salesperson adds, structured for machine retrieval.

QUERY EXPANSION

Fan out into your catalog

Live queries return curated follow-ups and store insights — suggested expansions with a pre-computed top pick — plus structured compare endpoints, so an agent goes deeper instead of bouncing.

AGENT SHORTCUTS

Built for how agents retrieve

Contextual links to your /agent/* endpoints and a potentialAction search hook collapse an agent’s multi-step scrape into a single direct query against your catalog.

Trace + intent graph

Every agent-driven visit compounds into your own intent graph.

Drop one component — <InceptionTraceScript /> — in your root layout. The beacon stitches the journey an agent started: the query it ran, the link a shopper clicked, the arrival on your store — correlated by a per-query ?ia= token.

Those arrivals feed a per-merchant intent graph that learns, with Thompson Sampling, which content wins for each platform and intent. The longer you run, the better you’re understood — and recommended.

  • Closed-loop attribution
    Every arrival traced back to the agent query that drove it — at your own edge, no Shopify required.
  • Per-merchant, compounding
    Your graph is yours. It sharpens with every interaction — a moat competitors can’t copy.
  • Privacy-first
    First-party signals from your own edge layer — no third-party cookies.
The compounding loop
Agent runs a query on your catalog
Shopper clicks through (?ia=)
Arrival stitched at your edge
Intent graph learns what wins
Recommended more often ↑
Composable, now agentic

Add the agentic layer that converts.

Keep your frontend, your framework, and your edge. Add the layer that makes your catalog discoverable, deeply explorable, and decision-ready for every major AI shopping agent — turning recommendations into conversions. One npm install, or one click.

inceptionagents.com