For Magento & Adobe Commerce agencies

Your brand, our protocol engineering.

Your clients are asking what ChatGPT, Google AI Mode and Perplexity see when they read their store. Answer under your own name. We do the feeds, edge policy, structured data and agent interfaces behind it, then hand the work to your developers.

Talk about a partnership Agency deck (PDF)

You
The client relationship, the brand on every deliverable, the invoice and the roadmap you sell after it.
You
Your developers, who build on the handover and keep the store afterwards.
Us
Reviewed scans, catalog and feed audit, agent-interface architecture, one working prototype, handover.
Us
Staying invisible to your client unless you decide to introduce us.

01 · Decide

Four things your agency can offer without building a new practice.

Each one can be sold alone. They are built to be delivered in order: a reviewed scan starts the conversation, an engagement settles it, and the answer pack and training mean your team can hold the conversation without us in the room.

WL-01

White-label scan and reviewed report

We run our eight public readiness checks against your client's live store, an engineer reviews the result and writes up the three risks that matter most, and the report goes out in your template, under your name.

Brand
Yours
Needs
Store URL and your client's permission
Access
None. Public pages only, read-only

WL-02

Engagement as your subcontractor

Catalog and feed audit, agent-interface architecture tied to the client's Magento, ERP, PIM and payment boundaries, one working prototype on an agreed path, and a production roadmap your team can quote.

Brand
Yours, or co-branded
Shape
Our standard sprint, or the two-week agency process below
Handover
To your developers

WL-03

ChatGPT Shopping answer pack

A short, sourced brief your account managers can send when a client asks "can we sell in ChatGPT?" It covers what is open today, what depends on programme approval, and what their store has to fix first.

Brand
Yours
Form
Client-ready brief and talking points
Kept current
When programmes or specs change

WL-04

Team training, 60–90 minutes

For your account managers and Magento developers: what shopping agents read, how the eight checks work, which fixes are theme work and which are integration work, and how to talk about it without overpromising.

Format
Live and remote; recording on request
For
Sales, account management, developers

How the commercials work

  • You buy from us and invoice your client at your own price.
  • Terms for each component are agreed with your agency in writing before the first engagement. They are not published here.
  • Every engagement starts with a written scope that both of us approve.
  • We do not contact your client unless you ask us to.
  • Reports and scan results stay private to you and your client.
  • Deliverables can be yours to rebrand and reuse for that client.
  • Nothing on this page is a binding offer until a scope is signed.

02 · Learn

What we actually check, and fix.

The same eight checks run on expandingweb.com today. They read public pages within a bounded budget. Blocked probes need review: a request declaring an AI user-agent is not a verified crawler. We distinguish missing evidence from a confirmed defect before proposing a fix.

  1. UCP-01Capability manifest at /.well-known/ucpCan an agent discover what the store supports?
  2. MCP-01MCP initialize handshakeIs there a live tool interface for catalog data?
  3. BOT-01robots.txt policy for AI crawlersWhich agents are allowed in, on purpose?
  4. AGENT-01How agent user-agents are served at the edgeDoes the WAF answer a shopping agent with a 403?
  5. CRAWL-01Crawl limits on search and faceted URLsIs the server protected from combinatorial URLs?
  6. LLMS-01llms.txtIs there a readable map of the store for models?
  7. FEED-01Product data completenessIdentifiers, shipping and returns fields agents need
  8. SCHEMA-01Product JSON-LD on the product pageCan the page quote a price and availability?

Prefer to scan first? Run the free public scan on your client's store.

03 · Learn

A two-week process, run alongside your team.

This is not the standard five-business-day Readiness Sprint sold on expandingweb.com. It is a longer shape for agencies whose developers stay in the loop. Dates, scope and terms are agreed per engagement.

Week one — evidence and design

Day 1
Scope call with your lead. Written scope approved by both sides.
Day 2
Reviewed scan and three prioritised risks, in your template.
Days 3–4
Catalog, feed and edge audit on the real store.
Day 5
Architecture for the agent interfaces, reviewed with your developers.

Week two — prototype and handover

Days 6–8
One working prototype on the agreed path, in your repository and staging.
Day 9
Handover session: code walkthrough, runbook, open risks.
Day 10
Your team presents the result and roadmap to the client.

04 · Learn

See the finished picture on our demonstration store.

We would rather show working artifacts than describe them, so we run one small store where every artifact the scanner checks exists for real.

ready.expandingweb.com

Nord Trail

A demonstration outdoor-gear catalog built and operated by Expanding Web. It serves a UCP manifest, a live MCP endpoint, an explicit AI-crawler policy, llms.txt, a sitemap and product JSON-LD from one catalog. Show it to your client as "this is what the finished layer looks like".

It is
Our own reference implementation, labelled as a demonstration on every page.
It is
Re-scanned daily by the same scanner your clients' stores go through.
It is not
A client, a case study or proof of revenue. Orders are not fulfilled.
It is not
A Magento installation. It is served by our own application, so it shows the target artifacts, not a Magento build of them.

05 · Learn

What neither of us will tell your client.

These limits are the reason a sceptical ecommerce director keeps reading. Put them in your own pitch.

No placement or ranking.

Discovery, ranking and what an assistant chooses to recommend stay with the platforms. We make a store readable and transactable; we do not promise where it appears.

No programme enrolment promises.

We prepare feeds and help with applications. Acceptance into Google, Perplexity or OpenAI merchant programmes is each programme's decision. Country eligibility is checked against current documentation.

No certification.

We do not issue an official "agent-ready" certification. A passing scan is a measurement on a date, not platform approval.

No borrowed proof.

No client logos, testimonials or uplift figures. The evidence is the client's own scan, run in front of them.

Who you work with

Practice
Expanding Web, founded 2007. 17+ years of Magento engineering, ERP, PIM, payments and B2B integrations.
Architecture sign-off
Marcin Dancewicz, founder. Magento Certified Developer Plus, former Klarna Solution Engineer.
Day-to-day contact
A named technical account manager for your agency, with written status at every phase.

Next step

Bring one client store. We'll start there.

Tell us about the agency and one client who has asked about AI shopping. We reply within two business days with a proposed scope for that one store. Terms follow the scope, not the other way round.

Talk about a partnership info@expandingweb.com