WebAnatomy

ai.webanatomyPublished by webanatomy.aiOfficial publisherv0.1.1Updated Oct 25, 2026

Help your AI improve landing pages, grounded in 3,500+ scored sections and 500 real pages.

VerifiedStreamable HTTPWeb executableData & AnalyticsBusiness & CommerceMedia & Design

Overview

AI-generated overview

A hosted MCP endpoint that gives an AI assistant live access to a scored library of real landing pages and sections for benchmarking, auditing and rewriting web

What you can do
Search and retrieve benchmark landing pages and individual sections from a scored library of real pages, so recommendations cite concrete examples instead of generic advice. Compare a page against top-ranked pages in its market and label how it stacks up. Support section-level research with tiered improvement guidance grounded in real benchmark sections. Feed grounded copywriting and page assembly workflows, including visual moodboards and shareable wireframes. The README notes the tools degrade to static guidance when the endpoint is not connected.
When to use it
Use it when an assistant is auditing, benchmarking, rewriting or building a landing page and you want its suggestions backed by real scored pages and reference screenshots rather than generic conversion advice. It pairs with the vendor's separate skill pack, which works without it but is described as best with it connected.
Requirements
A remote streamable HTTP endpoint; no local package. A WebAnatomy account and a personal bearer token generated from the vendor dashboard, sent in the Authorization header. The service is in private beta, so access must be requested first. Network access to the vendor endpoint.
Before you install
The Authorization header carries a personal bearer token; keep it secret and do not share it. The service is in private beta and the README marks its benchmark figures as numbers to confirm before release. Pages and sections you submit for comparison are sent to the vendor's hosted service. The README also describes a separate skill pack installed via npx or a plugin, which writes report files into your project.

Installation

In SourceWeft

  1. Open WebAnatomy in the dashboard and add it to a workspace.
  2. Connect Authorization token credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/ai.webanatomy%2Fwebanatomy/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "webanatomy": {
      "type": "http",
      "url": "https://mcp.webanatomy.ai/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_TOKEN>"
      }
    }
  }
}

README

Web Anatomy

Skills for your AI agent that ship landing pages that work. Built from the plays the best SaaS companies actually shipped, grounded in a scored benchmark library so every recommendation cites a real page instead of generic advice.

$ npx skills add GabrielAmz/web-anatomy

Installs the Web Anatomy skill pack for Claude Code, OpenAI Codex, Cursor, Windsurf, and any agent that supports the Agent Skills spec.

What Web Anatomy is

Most AI agents rework a page from generic advice: add social proof, make the CTA stronger, tighten the headline. Web Anatomy gives the agent evidence instead. The skills read a page as a structure of known sections (hero, value proposition, trust, testimonial, pricing, FAQ, CTA), then ground every move in a benchmark library of real pages scored section by section.

The benchmark behind the skills (numbers to confirm before release):

  • 3,500+ scored sections
  • 500+ landing pages
  • 290+ companies

The skills run standalone with built-in patterns. Connecting the hosted MCP (highly recommended) pulls the live benchmark data and real reference screenshots into every report. See Live benchmark data.

The four altitudes

Web Anatomy works at four zoom levels. The altitude matches the question being asked.

AltitudeThe questionSkill
PageWhat do the strongest pages in this market do, and how does mine compare?find-examples
AuditHow does my whole page stack up, and what should I change first?audit-page
SectionHow do I level up one section, from foundational to best-in-class?research-best-practices
FixWrite the grounded copy to rework a page or build a new one.write-page

After the copy is written, build-page assembles the structure + copy into a shareable wireframe with a coherent, single-exemplar design system. It is the terminal step on both branches (improve and create), so it sits above the altitudes rather than inside one.

Run webanatomy-setup once first. It captures product, ICP, industry, competitors, and proof assets into .agents/webanatomy-context.md, so every skill gives sharper, on-brand recommendations.

Two flows

  • Improve an existing page: audit-page (diagnose) -> write-page (rewrite the weak sections) -> build-page (optional: assemble the improved page as a shareable wireframe, in your page's own look).
  • Create a new page: find-examples (pick one exemplar homepage) -> write-page build mode (a light outline adopts the exemplar's section order, then writes the copy) -> build-page (assemble the wireframe in the exemplar's look).

The skills

SkillAltitudeWhat it does
webanatomy-setupFoundationRun once. Captures product, ICP, industry, competitors, proof assets, and priority pages into .agents/webanatomy-context.md.
find-examplesPageThe market scan. Pulls the top-ranked pages and sections in a market, shows what to steal, and when you share your page, labels how it compares. Absorbs the old benchmark-compare.
audit-pageAuditScores the current page against a CRO rubric, diagnoses it section by section, and returns a prioritized fix list. Runs standalone, no MCP required.
research-best-practicesSectionThe section deep-dive. Takes one section and returns a tiered improvement ladder (foundational, competitive, best-in-class), each grounded in real benchmark sections.
write-pageFixThe copy engine. Turns a diagnosis, a tier, or a chosen structure into grounded copy-paste sections; reworks an existing page or writes a new one.
build-pageAssembleThe terminal assembler. Takes a structure + its copy and lays them out as a shareable wireframe with a single-exemplar design system (a benchmark page for a new page, your own page when improving).
moodboardPageThe section-by-section visual board. Captures the live page section by section and pulls benchmarks from the MCP, then composes each section as current-state-large + problems + a benchmark screenshot grid + discipline-tagged practices + recommendation. Where find-examples shows what to steal, moodboard composes the whole page as a visual reference you build from.

Visual Reports

Web Anatomy does not rely on chat to render screenshots. The output skills write local visual artifacts:

txt
.webanatomy/├── find-examples/├── audit-page/├── research-best-practices/├── moodboard/│   └── page-date/│       ├── report.md│       ├── report.html│       └── references/├── write-page/│   └── topic-date/│       ├── report.md│       ├── report.html│       └── references/│           ├── current.png│           └── company-section.png└── build-page/    └── page-date/        ├── wireframe.html        ├── design-system.md        └── references/

The chat response should stay short: top findings, report path, and any screenshot or MCP limitations. report.html is the product experience.

Live benchmark data (optional, highly recommended)

The skills work on their own, so install them and start immediately. Connecting the hosted Web Anatomy MCP is never mandatory, but highly recommended: it grounds every recommendation in live benchmark data and real reference examples instead of static guidance, which is where the skills are at their best. It needs a beta token and one config command per IDE:

Sign in at https://www.webanatomy.ai/dashboard/mcp to get your token + per-IDE config. You must be signed in, because the page generates your personal token there.

The skills use MCP tools such as search_pages, get_page, search_sections, and get_section when connected, and degrade gracefully to static guidance when not. Scores, thresholds, raw marker coordinates, and benchmark field names stay internal; user-facing reports translate them into plain-English patterns and gap labels.

Want to see the benchmark first? Browse the public scored sections gallery at https://www.webanatomy.ai/best-landing-pages/sections — real landing-page sections, scored, with what makes each one work. It is the same library the MCP serves to your agent, so you can explore it before connecting anything.

Why "Anatomy"

Most AI-agent skill packs frame pages as a job (CRO, copywriting, SEO). Web Anatomy frames them as a structure: hero, proof, problem, solution, pricing, FAQ, CTA. Every section has known patterns, known anti-patterns, and known reference companies.

Every Web Anatomy skill turns a page into parts: hero, proof, problem, solution, pricing, FAQ, CTA. Each part has known patterns, anti-patterns, and benchmark examples. The framework and the evidence travel together.

Compatibility

  • ✅ Claude Code
  • ✅ OpenAI Codex
  • ✅ Cursor
  • ✅ Windsurf
  • ✅ Any agent that reads .agents/skills/<name>/SKILL.md per the Agent Skills spec

Install options

Option 1 — CLI (any agent):

bash
# Installs the whole pack to .agents/skills/npx skills add GabrielAmz/web-anatomy
# Update later (or just re-run the command above, it's idempotent)npx skills update GabrielAmz/web-anatomy

Option 2 — Claude Code plugin (version-aware updates):

text
/plugin marketplace add GabrielAmz/web-anatomy/plugin install web-anatomy

Claude Code knows the installed version and when a newer one ships. Update with /plugin marketplace update web-anatomy then /plugin update web-anatomy, or turn on auto-update for the marketplace to make it hands-off. Skills install namespaced: /web-anatomy:write-page, /web-anatomy:build-page, and so on.

Option 3 — clone and copy manually:

bash
git clone https://github.com/GabrielAmz/web-anatomycp -r web-anatomy/skills/* .agents/skills/

After install, start with webanatomy-setup. The benchmark-backed workflows degrade gracefully when MCP isn't connected, falling back to static/reference guidance, but they are at their best with live benchmark data, so connecting MCP is highly recommended.

Repo layout

skills/ holds the skill pack, the source of truth installed by npx skills add.

Contributing

PRs welcome. See CONTRIBUTING.md for the new-skill recipe: kebab-case, YAML front matter, clear workflow names, and body under 5K tokens.

License

MIT. Use it, fork it, sell it, ship it.

Built by Gabriel Amzallag. The Web Anatomy pack is a public extraction of the patterns powering a private landing-page benchmark and audit tool. The skills give you the framework; the analyzer gives you the data.

Source: README.md at commit 0374f8a

Tools

0
This server asks for a credential such as an API key, so SourceWeft could not read its tools without one and has not indexed them.

Version history

1
  1. v0.1.1LatestSep 16, 2026

Score

85% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts