How to Generate Ecommerce Product Photos in Claude with MCP

Updated October 2026 | Reading time: 12 min

TL;DR: Claude does not generate images on its own. It acts as a creative director that talks to an external image service through the Model Context Protocol (MCP). By connecting the Avocado MCP server to Claude, you can generate product packshots, lifestyle scenes, and ad-ready crops directly in a chat. This guide covers setup, prompt patterns, shot types, and a comparison to standalone product photography tools.

Jump to:

• Why Claude needs MCP for product photos
• Setup: connect Avocado MCP to Claude
• The product photo workflow in 5 steps
• Shot types and prompt templates
• Prompt engineering tips for product accuracy
• Claude MCP vs standalone product photography tools
• FAQ

Why Claude needs MCP for product photos

Claude can analyze images, plan shot lists, and write detailed prompts. It cannot render pixels. To generate an actual product photo, Claude needs to call an external image generation service.

The Model Context Protocol (MCP) is the bridge. It lets Claude send structured requests to a connected image service, track job progress, and retrieve the finished image. Think of Claude as the creative director and the MCP server as the studio.

For ecommerce teams, this setup has a practical advantage: Claude handles the conversation, prompt refinement, and job orchestration in one place. You describe the product, Claude structures the request, and the connected service produces the image.

Avocado MCP

Avocado AI

Generate images and video in Claude

Connect the Avocado MCP to Claude, Codex or Cursor and make ads from chat.

Set up the MCP

Setup: connect Avocado MCP to Claude

The Avocado MCP server works with Claude Desktop, Claude Code, and any MCP-compatible client. Here is the setup for Claude Desktop:

1. Get your API key. Sign in at avocadoai.co and go to your account settings. Generate an avk_ API key. Every paid plan includes MCP access.

2. Add the MCP server to Claude Desktop. Open Claude Desktop settings, navigate to the MCP servers section, and add:

• Server URL: https://www.avocadoai.co/mcp
• Authentication: Your avk_ API key

3. Verify the connection. Ask Claude: "List the available image generation models." If the MCP connection is working, Claude will call the models_list tool and return the current catalog.

4. Check your credit balance. Ask Claude: "Check my Avocado credit balance." This confirms both read and write access are working.

Once connected, Claude can generate images, check job status, and retrieve results through the MCP tools without you leaving the chat.

The product photo workflow in 5 steps

Step 1: Upload your product reference

Start by giving Claude your product photo. A clean, evenly lit pack shot on a neutral background works best. If you have multiple angles (front, side, back, top), upload all of them.

Claude will inventory what it can see: shape, materials, color, label placement, and packaging details. It should flag what it cannot determine from the reference alone.

"Here are reference photos of my vitamin bottle. What can you see, and what details are unclear?"

Step 2: Define the shot list

Tell Claude what images you need. Be specific about aspect ratio, background, lighting, and placement. "Make it premium" is not a brief.

Good example:

"I need 4 images for my Shopify listing:

1. Main packshot: 1:1, white background, centered, soft shadow

2. Lifestyle: 4:5, bathroom counter with natural morning light

3. Detail: tight crop on the label and cap

4. Scale: held in hand for size reference"

Step 3: Let Claude structure the prompt

Claude will take your brief and the reference photo details and build a structured prompt for the image model. It should include:

• The exact product description (from your reference)
• Background, lighting, and mood
• Aspect ratio and crop
• What must NOT change (label text, packaging shape, color)
• Negative prompts (no text, no extra objects, no brand names)

Step 4: Generate and review

Claude calls the generate_image tool through MCP, which returns a job ID. Claude then polls check_job until the image is ready. The result includes a URL to the generated image.

Review the output against your actual product. Check:

• Is the label text accurate?
• Does the packaging shape match?
• Is the color true to the real product?
• Are the proportions correct?

If something is off, tell Claude what needs to change. It will adjust the prompt and regenerate.

Step 5: Download and use

Once approved, download the image directly from the URL provided. The image is yours to use commercially with no watermarks on any Avocado plan.

Shot types and prompt templates

Here are five shot types every ecommerce brand needs, with prompt patterns that work well through Claude MCP.

1. Main packshot (marketplace listing)

Clean white background, centered product, soft grounding shadow. This is your Amazon, Shopify, or marketplace main image.

Pattern:

"Product packshot of [product description] on a pure white background. Centered, large in frame, soft contact shadow beneath. Even studio lighting, no harsh highlights. No text, no props, no hands."

2. Lifestyle scene

Product in context, showing how and where it is used. Lifestyle images help shoppers visualize the product in their own space.

Pattern:

"Lifestyle product photo of [product] on [surface/location]. [Natural/morning/evening] light from [direction]. [1-2 props that make sense, no clutter]. Shot from [angle] with [lens feel]."

3. Detail or texture shot

Tight crop on the label, material, or key feature. Useful for showing quality, ingredients, or craftsmanship.

Pattern:

"Close-up detail shot of [specific part: label, cap, stitching, texture]. Sharp focus on [detail]. Soft blurred background. Macro photography feel."

4. Scale reference

Product held in hand or placed next to a common object to communicate size.

Pattern:

"Person holding [product] in one hand at [angle]. Clean, slightly blurred background. Natural daylight. The product should be clearly visible and properly sized."

5. Flat lay (top-down)

Product arranged with complementary items, shot from directly above. Popular for social media and brand imagery.

Pattern:

"Overhead flat lay of [product] arranged with [2-3 complementary items]. Clean surface, even lighting, no harsh shadows. Generous negative space for text overlay."

Prompt engineering tips for product accuracy

Getting a good product photo through Claude MCP comes down to how well you protect the product truth and direct the creative.

Name the product exactly. "A 500ml amber glass dropper bottle with a matte black cap and a white label reading 'Botanical Glow Serum'" is better than "my serum bottle."

List what must not change. Claude will pass these as constraints to the image model. Label text, packaging color, cap style, and product shape should all be listed as invariants.

Specify the camera, not just the mood. "Shot at eye level with a 50mm lens, f/2.8, soft window light from the left" gives the model more to work with than "professional looking."

Use one product per generation. Group shots are harder to control. Generate individual products and composite them if needed.

Start with the hardest shot. If your product has a reflective surface, intricate label, or unusual shape, generate that shot first. If the model handles it well, simpler shots will follow.

Claude MCP vs standalone product photography tools

Claude MCP is a workflow, not a dedicated product photography tool. Here is how it compares to purpose-built alternatives.

Where standalone tools lead: Dedicated product photography tools like Photoroom and Claid offer specialized features that Claude MCP does not replicate: automatic background removal from a phone photo, batch processing of hundreds of SKUs, marketplace-specific compliance templates, and one-click ghost mannequin effects. If your only need is fast batch editing of existing product photos, a specialist tool is more efficient.

Where Claude MCP leads: The value is consolidation. Claude can generate a product packshot, then the same system creates a video ad, writes the caption, generates background music, and schedules the post. On the Avocado AI Starter plan at EUR 39/month, you get 300 credits across image, video, audio, and Flows. A one-credit model like Nano Banana 2 produces one image for roughly EUR 0.13. That same credit balance covers video, voiceovers, and music when you need them.

The honest trade-off: Claude MCP requires more setup and prompt skill than a drag-and-drop app. You are building a relationship with the tool, not just pressing a button. But for brands that need more than just product photos, the all-in-one system saves time and budget compared to stitching together 3-4 separate subscriptions.

How Avocado fits

Avocado AI is not a dedicated product photography tool. It is a creative system where Studio handles image, video, UGC, and audio generation in one workspace. The MCP server lets Claude, Codex, or Cursor use the same tools your browser does.

For product photos specifically, Avocado's image workbench supports multiple models suited to ecommerce: GPT Image 2.5 Flare for editorial-quality product scenes, Nano Banana 2 for fast packshots at 1 credit each, and Recraft v4.1 for design-focused layouts. You choose the model per shot type.

When you need to go beyond still images, the same system handles video product ads through Seedance 2.5, voiceovers through Audio Studio, and publishing through Super Agent on Growth and Pro plans.

FAQ

Can Claude generate product photos directly?

No. Claude can analyze uploaded images and write detailed prompts, but it does not render pixels. To generate actual product photos, Claude needs to call an external image service through MCP (Model Context Protocol). The Avocado MCP server connects Claude to models like GPT Image 2.5 and Nano Banana 2 for this purpose.

What product reference photos should I provide?

Upload a clean front or three-quarter pack shot in even, neutral light. Add side, back, top, and detail views if the product has labeling, textures, or features that change by angle. The more Claude can see, the more accurately it can describe the product to the image model.

How much does it cost to generate product photos through Claude MCP?

On Avocado AI, a single image with Nano Banana 2 at 2K resolution costs 1 credit. On the Starter plan (EUR 39/month for 300 credits), that works out to roughly EUR 0.13 per image. Higher-quality models like GPT Image 2.5 Flare use more credits. You see the exact cost before every generation.

Can I use AI-generated product photos on Amazon or Shopify?

That depends on the marketplace rules. Amazon requires that main images accurately represent the product and meet specific background and framing guidelines. Always review AI-generated images against the actual SKU before uploading. The image should be truthful, not just attractive.

How do I keep product labels accurate in generated images?

List label text and placement as invariants in your prompt. Claude will include these as constraints. If the model still alters the text, you can use the image edit feature to fix specific areas, or composite the real label onto the generated background in post-production.

What is the difference between Claude MCP and using Avocado's web app directly?

The web app gives you a visual workspace with presets, drag-and-drop editing, and direct model selection. Claude MCP gives you a conversational interface where Claude handles the prompting, model selection, and job tracking. Both use the same underlying models and credit balance. MCP is better for iterative refinement and multi-step workflows; the web app is better for visual exploration and batch work.

Do I need a paid Avocado plan to use the MCP server?

Yes. Every paid plan (Intro at EUR 19.99/month and above) includes MCP access. There is no free tier. You need an avk_ API key from your Avocado account to authenticate the MCP connection.

How to pick in under 30 seconds

• You want Claude to handle the full product photo workflow: connect the Avocado MCP server and describe your shots in chat.
• You only need fast background removal on phone photos: use Photoroom or Pixelcut instead.
• You want art-directed campaign scenes with 3D props: try Flair AI.
• You need to process hundreds of SKUs through an API: Claid AI or Nightjar are better fits.
• You want product photos plus video ads, voiceovers, and publishing from one system: Avocado AI covers all of it.
• You are already using Claude for other tasks: MCP lets you generate images without switching tools.
• You need marketplace-compliant packshots with strict accuracy: review every generated image against the real product before uploading.
• You want the lowest per-image cost: Nano Banana 2 on Avocado Starter is roughly EUR 0.13 per image.

If you want one creative system for ecommerce product photos, video ads, and content publishing, start with Avocado AI. Check out our pricing for details. The MCP server connects directly to Claude, so you can generate, iterate, and ship without leaving your chat.

Wanderson Jackson is the founder of Avocado AI, a creative production system for ecommerce brands. He writes about AI-powered content workflows for DTC and agency teams.