The AdFury.ai MCP Server
Model Context Protocol access to the Retail Media App and the Digital Shelf App, so the AI agents your team already works in can generate, resize, and publish compliant retail content by asking.
Your AI client
adfury-mcp connected“Take the Heirloom Manor cut crystal glassware listing, build the full Walmart ad set and the product page imagery, and keep everything inside our brand guidelines.”
- retail_media.creative.generate9 placements built from the product link, layered and editableRetail Media App
- retail_media.ads.resizeOutpainted to every Walmart Connect size, no cropsRetail Media App
- digital_shelf.images.generate14 PDP images: studio, lifestyle, and mobile infographicDigital Shelf App
- digital_shelf.listings.updateTitle, bullets, and A+ copy staged for approvalDigital Shelf App
Every call runs under your brand profile, retailer specs, and seat permissions — the same rules the apps enforce.
Request MCP AccessOne server, any MCP client your team already works in
- Claude
- ChatGPT
- Gemini
- Microsoft Copilot
- Cursor
- VS Code
- Claude Code
- Your own agents
The Model Context Protocol is an open standard, so anything that speaks it can reach AdFury. There is no AdFury-specific SDK to adopt and no plugin to maintain per assistant.
What MCP is
The protocol that lets an agent actually use AdFury
A model that can only describe retail work is a research assistant. The Model Context Protocol is what turns it into an operator: a common interface for calling real tools against real data, adopted across the industry so one server reaches every assistant.
- One connection, not one integration per assistantMCP is the open standard that lets an AI application talk to outside tools and data — the USB-C port of the agent world. Connect the AdFury server once and every client on it inherits the same capability, instead of commissioning a bespoke build for each assistant your organization adopts.
- Tools the agent discovers on its ownThe server advertises what it can do, so the model reads the available tools, picks the right one for the request, and chains several together when a task needs it. Nobody has to memorize a module name or learn where a setting lives in the interface.
- Plain language in, finished retail work outAsk for a Walmart ad set, a refreshed product page, or last week's creative performance and it happens in the thread. The output is real work in your AdFury workspace — layered, editable, and spec-checked — not a screenshot or a suggestion to go do it yourself.
- Governed the same way the apps areAuthentication is scoped to the user who connected, brand rules and retailer specs are enforced server-side, and every call is logged. An agent operating through MCP has exactly the reach the person behind it already had, and no more.
- MCP clientClaude · ChatGPT · Gemini · Cursor · your own agentYour side
- Model Context ProtocolOpen standard · tool discovery · OAuth · remote transportStandard
- AdFury.ai MCP serverRetail Media App tools · Digital Shelf App tools · brand policyAdFury
- Retail surfacesAmazon · Walmart · Instacart · 40+ ingested platformsYour estate
1Ask
A request in plain language
2Select
The agent picks the AdFury tools it needs
3Run
Generation happens under your brand profile
4Return
Assets, copy, and metrics land in the thread
Product data stays in your AdFury workspace. The agent calls tools against it rather than dragging the catalog through a context window.
What the server exposes
Two apps, one toolset
The AdFury MCP server publishes the Retail Media App and the Digital Shelf App as tools side by side. An agent can generate the ad and fix the product page it points to in the same conversation, because both run on the same catalog and the same brand profile.
Retail Media App
Campaign creative an agent can produce end to end
Everything the six modules do inside the app — creative, content, campaigns, insights, resizing, and the brand control center — is reachable as a tool call, so an agent can carry a campaign from product link to live placement without a human opening the interface.
- creative.generate
- Build every placement for an item from its product link, layered and editable, with retailer specs applied during generation rather than checked afterwards.
- content.keywords
- Pull live retailer keyword data, category competitors, and brand velocity so the agent writes copy aimed at demand that actually converts.
- content.generate_copy
- Return headlines, sub-headlines, and body copy for any item, written against the keywords the same call surfaced.
- ads.resize
- Outpaint an approved creative into every remaining size for the networks you sell on, reframing instead of cropping and staying fully editable.
- campaigns.list / campaigns.launch
- Read the state of every live campaign across networks and push a new one live, from a single platform-agnostic view.
- insights.query
- Answer performance questions in the thread — conversions, creative variants, blended ROAS — across the forty-plus platforms in one reporting view.
Digital Shelf App
Catalog content an agent can keep current
Imagery, listing content, and rich media for the whole catalog, exposed as tools an agent can run one SKU at a time or across thousands. The ad the network serves and the product page it lands on stay in step because both come from the same workspace.
- images.generate
- Produce a full set of layered product page imagery from a product URL: studio, lifestyle, and mobile-optimized infographics, no photoshoot involved.
- listings.search
- Find any organic listing you own across retailers and read back its current title, bullets, description, and imagery.
- listings.update
- Rewrite titles, descriptions, and images one at a time or in bulk across hundreds of SKUs, staged for approval before anything publishes.
- rich_media.publish
- Turn a live product page into an immersive one using modules like comparison tables, 360 viewers, and before-and-after sliders, published compliant.
- catalog.sync
- Trigger or check the retailer connections that pull product data in automatically, so the agent is always reasoning over the current catalog.
- shelf.health_score
- Score product detail pages on content completeness and surface the specific gaps worth fixing next.
In practice
Things worth asking once AdFury is on the other end
No syntax to learn and no module to open. Describe the retail outcome you want and the agent works out which AdFury tools get there.
“Build the full Walmart Connect ad set for our three best-selling SKUs and pull the keywords first.”
Live keyword data, copy written against it, and every required placement generated on brand in one pass.
“Legal approved the hero creative. Resize it for every network we run.”
One approved asset outpainted into each remaining size, reframed rather than cropped, still editable.
“Which creative variants beat their category benchmark last week, and where are we overspending?”
A performance read across every connected network, answered in the thread instead of in a dashboard.
“This product page only has two images. Give it a complete set.”
Studio, lifestyle, and mobile infographic imagery generated from the listing, ready for approval.
“Update every listing in the summer line with the new claim language and flag anything non-compliant.”
A bulk edit staged across hundreds of SKUs, with the listings that break policy separated out.
“We're launching Thursday. Get the ads and the product pages ready together.”
Campaign creative and detail page content produced from the same catalog record, so the click and the landing match.
Trust and control
An agent with access, not a blank cheque
Opening a workspace to autonomous software is only reasonable if the guardrails sit on the server. The controls that make the Retail Media App and Digital Shelf App safe for a team make them safe for an agent, because they are the same controls.
Scoped, per-user authorization
Connecting the server authenticates the individual, not the assistant. An agent inherits that person's seat, their brands, and nothing else.
Brand rules applied server-side
The brand profile built from your website — logo, palette, type, voice — plus your guidelines and legal policies govern every generated asset, whether a person or an agent asked for it.
Retailer specs enforced during generation
Placement dimensions and platform rules are applied while the asset is being made instead of caught in review, which is what keeps the rejection rate under one percent.
Brand data stays siloed
Every user sits under one account with view and edit rights set per brand. An agent working on one brand cannot read another's catalog, assets, or spend.
A reviewable trail of every call
Tool calls are logged against the user who authorized them, so there is a record of what an agent generated, changed, or published.
Publishing stays a decision
Generation and bulk edits can be staged for human approval, so an agent can prepare a launch without being the one that ships it.
Agent connection request
Northwind Brands · 1 seat- catalog:readGranted
- creative:generateGranted
- listings:writeApproval required
- campaigns:launchApproval required
- billing:readDenied
Two-factor authentication required before the first tool call
Getting connected
Three steps, no custom development
Point your client at the server
The AdFury server is remote and hosted, so there is nothing to install or keep running. Add the endpoint to Claude, ChatGPT, Gemini, Cursor, or your own agent the same way you add any other MCP server.
Authorize the seat
Sign in once and approve the scopes the agent may use. It picks up the brand profile, catalog connections, and permissions already configured in your workspace.
Start asking
The tools announce themselves, so the agent knows what the Retail Media App and Digital Shelf App can do from the first message. Ask for the work and review what comes back.
MCP
Frequently asked questions
Put retail production inside the agent you already use
One MCP connection hands your AI the Retail Media App and the Digital Shelf App, governed by the brand rules you already set.
Why AdFury
The AdFury difference
Retail expertise. Purpose-built AI for retail. Operational excellence. Customer accountability.
The Retail AI Brain
Building our AI infrastructure since Oct. 2024. Agentic App. Managed Services. Custom AI workflows. Purpose-built for retail, not adapted from generic AI.
Retail Insiders, Real Results
Built by founders who scaled retail SaaS to exit. Deep expertise in retail media networks and retail product content management. We know the problems.
Knowledge Meets AI Speed
Retail knowledge powered by fine-tuned AI. Adapt faster. Solve better. Evolve precisely. Momentum catapults results. Trust expertise.
Customer Success, Always
Specialists, not generalists. We go deep in retail for focus on growth. Proven founders. Managed services. Dedicated support. We never fail our clients.
Our Values
What drives us forward
Seven principles that decide what we build, how we test it, and who we build it with.
See what each value means in practice- See the Future
- Never Stop Researching
- Never Stop Testing
- Be the First
- Explore Relevancy
- Bring Everyone Together
- Inspire Rapid Change
Built by Retail Insiders
Proven founders. Deep retail expertise.
Founded by retail technology entrepreneurs who built and scaled WhyteSpyder to a successful exit, AdFury.ai combines decades of retail and e-commerce expertise with cutting-edge AI architecture. We don't just understand retail — we've built and scaled profitable SaaS solutions and managed services within it.
WhyteSpyder started as a digital shopper agency for Walmart in 2012, became retail ecommerce SaaS in 2019, and sold to Ascential in 2021, where the team went on to scale retail media inside Flywheel Digital. AdFury.ai followed in October 2024.
- Founded
- Oct. 2024
- Headquarters
- Rogers, AR
- Built for
- Retail only
- Team
- 14 insiders
