Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/AgricIDaniel/claude-seo/llms.txt

Use this file to discover all available pages before exploring further.

The Banana extension brings AI image generation into Claude SEO via the Claude Banana Creative Director pipeline, powered by Google Gemini. Instead of passing a raw description to an image model, Claude acts as a Creative Director: it analyzes your intent, selects the right domain mode and aspect ratio for the use case, and constructs an optimized prompt using a 6-component Reasoning Brief system before generation. The result is production-ready SEO images — OG/social previews, blog hero images, product photography, and infographics — with correct dimensions, appropriate style, and a post-generation SEO checklist covering alt text, file naming, WebP conversion, and schema markup.

​
What Banana adds

  • OG/social preview images — 1200×630 output at 1K resolution, professional and text-friendly
  • Blog hero images — dramatic widescreen 16:9 at 2K resolution, editorial quality
  • Product photography — clean white background studio shots at 2K resolution
  • Infographics — vertical 2:3 layout at 4K resolution with structured data emphasis
  • Batch generation — generate N variations of any use case in a single command
  • Audit integration — during /seo audit, a dedicated agent scans existing OG images and produces a prioritized generation plan (never auto-generates)

​
Prerequisites

  • Claude SEO installed
  • Node.js 18+
  • A free Google AI API key from aistudio.google.com/apikey
  • ImageMagick (optional, for post-processing: WebP conversion, cropping, background removal)
The Google AI free tier provides approximately 10 requests per minute and 500 requests per day. For most SEO image workflows this is sufficient. Rate limits reset at midnight Pacific time.

​
Installation

1

Clone the repo (if you haven't already)

git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
cd claude-seo
2

Run the installer

./extensions/banana/install.sh
The installer (1) verifies Claude SEO is present, (2) prompts for your Google AI API key if the nanobanana-mcp server is not already configured, (3) installs the seo-image-gen skill and agent files, and (4) writes the MCP server entry to ~/.claude/settings.json.
3

Restart Claude Code

Close and reopen Claude Code so the nanobanana-mcp server is loaded.
4

Verify

/seo image-gen og "Professional SaaS dashboard screenshot"
The skill confirms MCP availability before generating. If tools are not found, it prints the install command.

​
Commands

CommandWhat it generatesAspect ratioResolutionApprox. cost
/seo image-gen og <description>OG/social preview image16:91K~$0.04
/seo image-gen hero <description>Blog hero image16:92K~$0.08
/seo image-gen product <description>Product photography (white BG)4:32K~$0.08
/seo image-gen infographic <description>Vertical infographic2:34K~$0.16
/seo image-gen custom <description>Custom with full Creative Director pipelineconfigurableconfigurablevaries
/seo image-gen batch <description> [N]N variations of a use case (default: 3)per use caseper use caseN × base cost

​
Examples

# OG preview for a blog post on keyword research
/seo image-gen og "SEO keyword research dashboard with data charts"

# Hero image for an article about Core Web Vitals
/seo image-gen hero "Web performance visualization, colorful speed metrics"

# Product shot for an e-commerce listing
/seo image-gen product "Blue ceramic coffee mug with handle, studio"

# Infographic about AI search optimization
/seo image-gen infographic "AI search optimization statistics 2025"

# 3 OG image variations to pick the best one
/seo image-gen batch "SaaS onboarding flow screenshot" 3

​
The Creative Director pipeline

Claude never passes your raw text directly to the image API. For every generation it:
  1. Identifies the use case and loads the matching preset (aspect ratio, resolution, domain mode)
  2. Constructs a Reasoning Brief using a 6-component system:
    • Subject (30%) — physical specificity and micro-details
    • Style (25%) — camera specs, film stock, brand references
    • Context (15%) — location, time, weather, supporting elements
    • Action (10%) — pose, gesture, movement, state
    • Composition (10%) — shot type, framing, focal length
    • Lighting (10%) — direction, quality, color temperature
  3. Generates via the gemini_generate_image MCP tool
  4. Returns the image with a full SEO checklist

​
Post-generation SEO checklist

After every successful generation, the skill provides:
  • Alt text — a keyword-rich, descriptive suggestion
  • File naming — SEO-friendly format: keyword-description-widthxheight.webp
  • WebP conversion command: 
    magick output.png -quality 85 output.webp
  • ImageObject schema snippet for structured data
  • OG meta tags markup for social preview images: 
    <meta property="og:image" content="https://example.com/images/og-image.webp" />
<meta property="og:image:width" content="1200" />
<meta property="og:image:height" content="630" />
<meta property="og:image:alt" content="Descriptive alt text" />

​
IPTC TrainedAlgorithmicMedia metadata

AI-generated product images used in Google Merchant Center feeds must carry IPTC TrainedAlgorithmicMedia metadata to comply with Google’s policy on AI-generated product imagery. This IPTC field signals to automated systems — including Google’s feed validator — that an image was produced by a machine-learning model, enabling proper policy enforcement and preventing listing disapprovals. The Banana extension’s post-processing pipeline can write this metadata tag via ImageMagick after generation: 
magick output.png -set "IPTC:2:101" "TrainedAlgorithmicMedia" product-final.png
When you run /seo ecommerce on a site that uses AI-generated product images, the audit checks for the presence of this tag and flags any product images missing it as a compliance issue for Google Merchant Center.
Google Merchant Center’s policy on AI-generated product imagery is enforced at the feed level. Missing TrainedAlgorithmicMedia metadata on AI-generated product images can cause listing disapprovals. Run /seo ecommerce <url> to audit your product image compliance.

​
Integration with /seo ecommerce

During a /seo ecommerce audit, the seo-image-gen agent (distinct from the interactive skill) runs as an analyst-only subagent. It:
  1. Audits all OG and social images across the site for quality, dimensions, and missing images
  2. Checks product images for TrainedAlgorithmicMedia IPTC compliance
  3. Produces a prioritized generation plan with prompt suggestions and cost estimates
  4. Does not auto-generate any images — the plan is for your review before you run /seo image-gen

​
Uninstall

./extensions/banana/uninstall.sh
This removes the seo-image-gen skill and agent. If you also use the standalone Claude Banana skill, the nanobanana-mcp server configuration in settings.json is preserved.

​
Troubleshooting

SymptomCauseFix
MCP tools not availableServer not loadedRestart Claude Code; verify ~/.claude/settings.json has the nanobanana-mcp entry
API key invalidWrong or revoked keyGenerate a new key at aistudio.google.com/apikey; re-run the installer
429 Rate LimitedFree tier RPM exceededWait 60 seconds and retry; free tier: ~10 RPM, ~500 RPD
IMAGE_SAFETY errorSafety filter triggered (often a false positive)Claude suggests rephrased alternatives automatically
Images not appearingOutput directory not foundDefault output dir: ~/Documents/nanobanana_generated/; check the path returned after generation

Build docs developers (and LLMs) love

Get started for freeTalk to us