Amazon Agent Iris

Generate and optimize Amazon listing images from live account data.

Amazon Agent Iris is an Amazon listing optimization tool for product images. It combines AI image generation with live Seller Central data from the Selling Partner MCP, Amazon Ads performance, and the Product Creative Skill that applies Amazon's image rules. Together they generate, audit, publish, and verify compliant creative from the AI client your team already uses.

Read the guide

Gemini Nano Banana

OpenAI gpt-image-2

Seller Central rules

The loop

How Amazon listing image optimization closes the loop.

Find. The Selling Partner MCP reads your live catalog and image state. Iris surfaces empty image slots, single-image listings, suppressions, and rule failures, then ranks the work so your team knows where to start.

Prioritize. The Amazon Ads MCP adds click and conversion signals. That context helps the agent choose which image to improve and which creative hypothesis to test instead of generating variants without a performance reason.

Generate or edit. Iris creates a compliant main image or a full stack of feature, infographic, lifestyle, and detail assets from the real product. It can also audit a rejected image and map the issue to the exact visual fix.

Publish and verify. After you approve an asset, the Selling Partner MCP submits it to the live listing and confirms Amazon ingested it. Nothing goes live until you say so.

Iris owns visual listing optimization. The Amazon Listing Optimizer handles titles, bullets, search terms, and A+ structure. Both can work from the same live listing through the MCPs.

Why us

Iris is image-focused, listing-aware, and connected to the actions around it.

Seller Central gives it listing context

The Selling Partner MCP supplies catalog attributes, the current image stack, category context, and suppression state from the live account.

Amazon Ads gives it performance context

Click and conversion signals help the agent decide which image to improve and what creative hypothesis to test next.

It builds the complete image stack

Generate or edit main, feature, infographic, lifestyle, and detail images from the real product instead of a generic prompt.

Fixes rejects, not just makes new ones

Audit any image for a pass / fix / reject verdict and the exact change — including the error-100239 title mismatch.

Your brand survives every edit

Logos, labels, and printed text reproduce exactly on every chained edit; a garbled label is treated as a failure.

The approved asset closes the loop

The Selling Partner MCP publishes the image after approval and verifies that Amazon ingested it. The workflow does not stop at a download.

Category comparison

How Iris differs from an Amazon listing image generator.

A standalone Amazon listing image generator starts with a public product page or an uploaded photo and stops at a downloadable file. Iris combines image generation with live Seller Central context, Amazon Ads performance, policy checks, approval-gated publishing, and verification, so each asset remains connected to the listing it is meant to improve.

Iris owns the visual work. Run it with the Product Creative Skill for main images, lifestyle scenes, infographics, detail shots, audits, and suppression fixes. For titles, bullets, search terms, and A+ structure, use the Amazon Listing Optimizer. For setup, see the Amazon Agent Iris documentation.

Capability	Standalone image generator	Amazon Agent Iris
Listing context	Public ASIN or uploaded photo	Live catalog attributes and current image stack
Performance signal	Generic templates or manual research	Amazon Ads click and conversion data
Image work	Generates downloadable assets	Generates, edits, and audits the complete image stack
Compliance	Template guardrails	Listing-aware rule check with exact fixes
Publishing	Manual upload	Approval-gated publish through the Selling Partner MCP
Verification	Stops at download	Confirms Amazon ingested the approved asset

No meter

No image credits. No generation caps.

The other tools meter you: credits per image, image counts per tier, a top-up when you run dry. Iris doesn't. You bring your own Gemini or OpenAI key, generate as many images as it allows, and pay the model provider directly, at their cost. Kuudo prices by team and marketplaces, not by how many images you make — and it runs in your own infrastructure, not ours.

What you get

Amazon listing image optimization across the full creative stack.

Built from your real product

Iris works from the live listing — the actual product, its attributes, its images. Not a scraped public page, not a blank prompt you have to guess at.

Build the stack that sells

Most listings stop at the main image. Generate the infographic, lifestyle, and detail shots too: the slots that win the click and fill the cart.

Change one thing, keep the rest

Tweak a color, swap a background, try a new angle. The product stays itself across every edit, so you never start from scratch.

Brand fidelity is a hard rule

Logos, labels, and printed text are held exact through every change — a garbled label is treated as a failure, not something that ships.

Every size you need, one call

Zoom-quality hero shots, mobile-first crops, and ad-ready dimensions, square through wide and up to 4K, all from a single request.

Audited before Amazon sees it

Current Seller Central rules are built into every image and every audit, so you get a clear pass / fix / reject verdict with the exact change — before you submit.

Real output

Amazon listing images generated by Iris.

Every one built from a live listing and checked against Amazon's rules — no stock, no mockups.

Lifestyle
Lifestyle scene, seeded from the product

What the agents say

session feedback

› What did Kuudo actually change about how you work? Be specific.

"This is the first time I've seen image generation and live marketplace publishing work as one motion. The Amazon Agent Iris MCP produced lifestyle and infographic images seeded from the real product, and the Selling Partner API MCP wrote them straight onto the listing and let me verify they ingested into Amazon's CDN — no manual uploads, no Seller Central tabs. Just as important were the guardrails: exact measurements pulled from the catalog, a policy validator gating every image, and the judgment calls left to a human. Powerful and careful at the same time."

Claude Anthropic's AI assistant

For Sellers · 3P · FBA · SMB

A rejected photo shouldn't take your listing out of search.

Jump to section

For Vendors · 1P · Brand owners

One rule set, applied across every ASIN.

Jump to section

For Advertisers · Sponsored Brands · SD · Agencies

Image stacks that win the click — as many variants as your tests need.

Jump to section

For Sellers · 3P · FBA · SMB

A rejected photo shouldn't take your listing out of search.

A rejected main image can pull your whole listing out of search. Ask your assistant for a replacement and it comes back in minutes, already checked against live Seller Central rules, with the one or two fixes named in plain English. You stop guessing why Amazon said no.

Explore image auditing in beta

A search-ready main image checked against pure-white background, about 85% product fill, and full-product visibility rules
Know why a photo would fail and the exact change to make — before Amazon ever sees it
Turn a suppression into a quick fix instead of a fire drill, with the issue → fix mapping (including error 100239) applied for you

For Vendors · 1P · Brand owners

One rule set, applied across every ASIN.

Bring every ASIN up to the same Amazon-defined image standard. Your brand labels, logos, tags, and printed text stay exact, the product keeps looking like itself through every edit, and finished assets drop straight into the PIM and DAM your team already runs.

Explore catalog workflows in beta

Hold one consistent look across the whole catalog, even as you scale to thousands of SKUs
Logos, labels, and printed text are held exact on every edit — a garbled label counts as a failure
Move assets in and out securely over signed URLs, with no key sprawl and nothing to clean up

For Advertisers · Sponsored Brands · SD · Agencies

Image stacks that win the click — as many variants as your tests need.

Creative is usually what holds a test back. Here you spin up lifestyle, infographic, and detail variants as fast as you can think of them — seeded from the real product, and checked against Amazon's content rules. Change one element, keep the rest, and the next variant is A/B-ready in minutes.

Explore merchandising in beta

Test more and test faster: lifestyle, infographic, and detail variants whenever a campaign needs them
Pair it with the Ads MCP to pull your ad performance in as a seed for what to make next
Every variant is checked against Amazon's content rules before it enters the test

How it works

How Amazon listing image optimization works in three steps.

step 01

Read the listing and performance signals

The Selling Partner MCP reads the live catalog and current image state. The Amazon Ads MCP brings click and conversion context so the agent can prioritize the listing and the creative hypothesis.

step 02

Generate and audit the image stack

Ask for a main, lifestyle, infographic, feature, or detail image. Iris builds it from the real product, preserves the brand, and checks the result against the rules that apply.

step 03

Approve, publish, and verify

You approve the asset before any live change. The Selling Partner MCP publishes it to the listing and confirms Amazon ingested it, so the workflow ends with a verified result instead of a file in a download folder.

The floor

Audit Amazon listing images against the rules that apply.

Point the agent at any image and get a line-by-line read — what passes, what needs changes, and the exact fix — before Amazon ever sees it. Amazon still makes the final call; you just stop being surprised by it.

audit · main_image_01.jpg skill · amazon-product-image

background · pass

white ~254 · within model tolerance · normalization recommended

product fill · pass

~84% of frame · within 85% rule

title match · needs changes

error 100239 risk · title mentions "2-pack" · image shows 1 unit

props / text · reject

"new!" badge on lower-left · remove before submission

verdict · needs changes · 2 fixes

Honest framing

We get you compliant fast. Amazon still makes the final call.

Acceptance

We de-risk and accelerate compliance. We don't guarantee Amazon's final approval — Amazon does.

White background

Image models render at ~254, not exact 255. We get you studio-clean white and flag when a deterministic normalization pass is needed.

One provider per session

Gemini or OpenAI, chosen by the key on the request. No lock-in — but not both in the same call.

Interface, not retouching

We're the compliance brain and the generation interface. The models render the pixels; we never pretend it's a manual studio service.

Frequently asked questions

The questions teams ask before they ship.

What is an Amazon listing optimization tool for images?

It identifies which listing images need work, uses live catalog and performance context to generate or edit the right assets, checks them against Amazon's rules, and closes the loop through approved publishing and verification. Iris does this through its connections to the Selling Partner and Amazon Ads MCPs.

How is Iris different from an Amazon listing image generator?

A standalone generator usually starts with a public product page or uploaded photo and stops at a downloadable file. Iris works from live listing context, uses Amazon Ads performance to inform what to test, audits the result, and can publish and verify an approved image through the Selling Partner MCP.

Can Iris use Amazon Ads performance to optimize listing images?

Yes. Paired with the Amazon Ads MCP, Iris can use click and conversion performance as context for which image to improve or what variant to test next. The performance signal guides the creative decision; Iris remains responsible for generating, editing, and auditing the image.

Can Iris create images for A+ Content?

Yes. Iris can generate and audit the image assets used in A+ Content. The Selling Partner MCP handles the A+ document and Amazon submission workflow, so image generation stays listing-aware without making Iris responsible for the entire content document.

Can Iris publish images directly to Seller Central?

Yes, after you approve them. The Selling Partner MCP submits the approved asset to the live listing and verifies that Amazon ingested it. Iris does not bypass your approval or claim that submission guarantees Amazon's final acceptance.

Will this guarantee my image gets approved?

No. Amazon makes the final acceptance decision. Iris applies the relevant rules during generation and audits the image before submission, which reduces avoidable rejections but does not replace Amazon's review.

Does it preserve my logo and printed text when editing?

Yes. Exact logo, label, and typography fidelity is a first-class rule on every chained edit. A garbled label is treated as a failure, not something that ships.

Gemini or OpenAI — which, and can I switch?

We recommend Google Nano Banana for speed and cost. However, you can use any Gemini or OpenAI model, per-session, chosen automatically from the API key the request carries. No lock-in. The compliance bar is identical on both.

Does it handle apparel and footwear model rules?

Yes — on-model vs flat by audience, kids and multipacks flat, footwear single shoe facing left at 45°, and the other category-specific rules from the skill.

What about multipacks, variants, and used-item offer photos?

Iris checks pack count, variant consistency, and offer-image requirements against the live listing. It applies the relevant special-case rules and names the exact fix when an asset conflicts with the listing.

Do you charge per image?

No. You bring your own Gemini or OpenAI key and pay the model provider directly at its cost. Kuudo does not meter image generations or sell image credits.

How do images get in and out securely?

MCP-native upload plus signed download URLs — links instead of shared keys. Configurable retention or ephemeral TTL on assets and links.

Where does it run?

In your own infrastructure — customer-hosted is the standard model. Docker-ready, works with any MCP client (Claude, ChatGPT, Cursor, IDEs, automation), and drops into the tools you already operate.

Private beta

Optimize Amazon listing images from the assistant you already use.

Explore the private beta Audit a listing image

Guided onboarding · works in any MCP client · runs in your own infrastructure