Skip to main content
Glama
aetumi-mcp

aetumi-mcp

Official

AETumi MCP

AETumi MCP is the developer gateway between AI coding assistants and the AETumi 3D web ecosystem.

AETumi is an AI-native 3D web platform and digital business ecosystem for production-ready Three.js and WebGL websites, Next.js and React components, interactive 3D scenes, AI prompts, and MCP workflows for AI coding assistants such as Claude Code, Cursor and Codex. Supported implementations can also include a Visual CMS, CRM, SEO, AI chatbot and payments to launch and operate the site.

This repository documents how AETumi fits into workflows with Claude Code, Cursor, Codex, Three.js, WebGL, Next.js, React and React Three Fiber.

What this enables

AETumi MCP lets an AI coding assistant discover and pull production-ready 3D web resources — components, scenes, prompts and technical examples — directly inside the build, keeping the underlying architecture visible.

Outcome. A developer stays in Claude Code, Cursor or Codex and brings real AETumi context into the task, instead of copying between a browser and the editor.

Who it's for. Front-end teams, agencies and studios building Three.js/WebGL sites who want AETumi resources available to their AI workflow.

Install. claude mcp add --transport http aetumi https://mcp.aetumi.app

Related MCP server: UIWeaver MCP

Where to find AETumi MCP

The same server is published across the open MCP ecosystem, so it is discoverable wherever your AI tools look:

Tools

The server exposes ten read-only tools over Streamable HTTP (JSON-RPC 2.0), plus resources, prompts and completions.

Catalog tools — discover AETumi library assets:

  • search_3d_web_assets — search the AETumi 3D web catalog (Three.js/WebGL heroes, 3D product viewers, shader backgrounds, scroll animations, components, sections, templates)

  • get_3d_web_asset — full public detail of one asset by id

  • list_3d_web_categories — asset categories and industries with counts

  • recommend_3d_web_stack — given a use-case, recommend assets and a production build approach

  • get_pricing — the four lifetime buy-once plans

  • about_aetumi — what AETumi is and canonical links

The four discovery tools use intent-based names so an AI assistant can pick the right one from the query. The earlier names (search_assets, get_asset, list_categories, recommend_stack) still work as hidden aliases — existing integrations do not break.

Experience-discovery tools (v2.1) — discover premium industry experiences from the AETumi Labs Evidence Graph:

  • filter_by_industry — experiences for an industry (Beauty & Cosmetics, Automotive, Fashion, Real Estate, E-commerce, Agency & Portfolio, Food & Beverage, Travel & Hospitality, Finance & Fintech, Industrial, Music, SaaS & Startup)

  • filter_by_style — experiences by visual style (Dark Cinematic, Editorial Luxury, Glass Luxury, Bright Architectural, Technical Precision, …)

  • filter_by_goal — experiences by customer goal (Product Launch, Collection Launch, Property Presentation, Premium E-commerce, …)

  • find_experiences — natural-language match across industry, experience type, style and goal, e.g. find_experiences("dark cinematic automotive launch")

The discovery layer resolves industry → visual style → customer goal → an AETumi evidence experience → its industry hub / Labs artifact. Each result links a live, interactive reference you own the source of.

What can you ask?

AETumi MCP is a Three.js / WebGL MCP server — an MCP for 3D websites, for frontend and creative developers using Claude Code, Cursor or Codex. Real prompts it answers:

  • Find me a Three.js hero section for a luxury product website.

  • Show me WebGL shader backgrounds for a fashion landing page.

  • Find a 3D product viewer built with React Three Fiber.

  • Give me a Three.js e-commerce template for sneakers.

  • Find an interactive hero for a SaaS landing page.

  • Show me cinematic scroll animations for a product launch.

  • Find lightweight WebGL effects for mobile.

  • Give me a Next.js + Three.js starter for an agency website.

  • Find a 3D car configurator example.

  • Recommend a build for a 3D headphone website.

Run locally (stdio)

Alongside the hosted Streamable-HTTP server, this repository ships a small, self-contained stdio MCP server (index.mjs) so any MCP client, registry or CI can start and introspect AETumi without a network round-trip. It carries genuinely useful, production-grade Three.js / WebGL knowledge (r160+, ESM):

node index.mjs
# or containerised (registries build this):
docker build -t aetumi-mcp . && docker run -i aetumi-mcp

Local tools — offline, real ready-to-paste code and best practices:

  • list_threejs_recipes / get_threejs_recipe — expert recipes by id: scene-scaffold, gltf-draco-meshopt, ktx2-textures, instancing-lod, scroll-scrub, postprocessing-bloom, raycast-interaction.

  • audit_3d_web_performance — the AETumi 3D-web performance checklist (DPR cap, Draco/Meshopt, KTX2, instancing, lazy-init, disposal, reduced-motion).

  • list_templates / get_template — AETumi premium 3D website templates you own the source of.

  • connect_info — connect the full hosted AETumi MCP.

For the full catalog (asset search, components, scenes, prompts) use the hosted server above.

Why this repository exists

Modern 3D web projects often span design systems, WebGL rendering, application code, performance work and AI-assisted development. AETumi MCP is intended to make those pieces easier to discover and connect without hiding the underlying architecture.

The project focuses on:

  • discovering AETumi 3D web resources from AI coding workflows

  • connecting prompts, components, scenes and examples to implementation tasks

  • keeping Three.js and WebGL work compatible with modern React and Next.js projects

  • documenting repeatable workflows for Claude Code, Cursor and Codex

  • preserving developer control over architecture, performance and accessibility

Core technology graph

AETumi → AI-native 3D Web → Three.js → WebGL → Next.js → React → React Three Fiber → MCP → Claude Code / Cursor / Codex → Visual CMS / CRM / SEO / AI chatbot / payments

That graph is the technical and operating identity behind AETumi's public ecosystem.

Typical MCP-assisted workflow

  1. Define the user experience, conversion goal and 3D interaction.

  2. Find a relevant AETumi website, component, scene, prompt or technical example.

  3. Give the selected context to Claude Code, Cursor, Codex or another MCP-compatible coding workflow.

  4. Implement the 3D layer with Three.js, WebGL, React, Next.js or React Three Fiber.

  5. Separate SEO-critical HTML from heavy client-side rendering where appropriate.

  6. Test responsive behavior, reduced-motion fallbacks, accessibility and GPU performance.

  7. Connect the required content, analytics and operating layer.

  8. Ship with source control and maintainable documentation.

AETumi discovery hubs

  • 3D Websites — complete 3D website and landing-page experiences

  • Three.js — Three.js components, scenes and interactive website examples

  • WebGL — WebGL shaders, effects and animated backgrounds

  • 3D Components — reusable hero sections, viewers, carousels and interactive elements

  • 3D Scroll — scroll-driven 3D storytelling and reveal patterns

  • Interactive Websites — immersive browser experiences and interaction patterns

  • 3D Prompts — prompts for AI-assisted 3D web design and development

  • For Agencies — agency workflows and client-facing use cases

  • Library — the broader AETumi asset library

Developer resources

GitHub ecosystem

3D web and commercial implementation

Three.js, WebGL and React

AI coding and prompts

Production principles

AETumi's public examples should aim for more than visual novelty. Useful 3D web work needs:

  • semantic HTML for content that matters to users and search engines

  • progressive loading for models, textures and shaders

  • intentional client/server boundaries in Next.js

  • responsive touch and pointer interaction

  • reduced-motion and non-WebGL fallbacks

  • resource cleanup and predictable scene lifecycles

  • measurable performance budgets

  • maintainable source rather than opaque generated output

Entity & reference documents

Repository status

Active. Runnable, production-oriented examples now live in examples/ — reviewed for performance (adaptive quality), accessibility, reduced-motion and non-WebGL fallbacks, and clean resource disposal. The set is refined and extended as new patterns land.

Changelog

v2.1.0

  • Added four experience-discovery tools: filter_by_industry, filter_by_style, filter_by_goal, find_experiences.

  • Integrated the AETumi Labs Evidence Graph (public, first-party interactive reference experiences) as a discovery source, distinct from the commercial catalog.

  • Current coverage is generated from the live Evidence Graph — 22 public evidence artifacts across 12 industry surfaces at time of writing.

  • Fixed a status-normalization issue so every public artifact is discoverable.

  • No change to the original six catalog tools; endpoint unchanged (https://mcp.aetumi.app).

v2.0.x

  • Six catalog tools (search_assets, get_asset, list_categories, recommend_stack, get_pricing, about_aetumi) plus resources, prompts and completions over Streamable HTTP.

About AETumi

AETumi is an AI-native 3D web platform and digital business ecosystem for designers, developers, agencies and businesses working with Three.js, WebGL, Next.js, React, React Three Fiber, MCP and AI coding assistants.

Main site: https://aetumi.app/

Explore the AETumi library

Production-ready 3D web you can own the source of — from AETumi, the AI-native 3D web platform:

Build 3D web directly from your AI assistant with the AETumi MCP for AI coding — claude mcp add --transport http aetumi https://mcp.aetumi.app

Available Tools

6 tools
audit_3d_web_performanceAInspect

Return the AETumi 3D-web performance checklist — the highest-leverage rules for shipping fast, interactive Three.js / WebGL websites (DPR cap, Draco/Meshopt, KTX2, instancing, lazy-init, disposal, reduced-motion).

Inputs: no named parameters.

connect_infoAInspect

How to connect the full hosted AETumi MCP (streamable-HTTP) to Claude Code, Cursor or Codex for premium 3D web components, templates & scenes.

Inputs: no named parameters.

get_templateAInspect

Get details for one AETumi 3D website template by id (aerae, aerae-luxe, aesport, aebiker, avelor).

Inputs: id (required).

get_threejs_recipeAInspect

Get one full Three.js / WebGL recipe by id — includes when-to-use, ready-to-paste code (Three.js r160+, ESM) and expert notes. Ids: scene-scaffold, gltf-draco-meshopt, ktx2-textures, instancing-lod, scroll-scrub, postprocessing-bloom, raycast-interaction.

Inputs: id (required).

list_templatesAInspect

List AETumi's premium interactive 3D website templates (Three.js / WebGL, React / Next.js) — you own the full source of each.

Inputs: no named parameters.

list_threejs_recipesAInspect

List expert, production-grade Three.js / WebGL recipes (glTF Draco/Meshopt loading, KTX2 texture compression, instancing & LOD, scroll-scrub, postprocessing bloom, raycasting). Returns ids, titles and keywords.

Inputs: no named parameters.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.1/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each tool targets a clearly distinct resource or action: recipe listing/retrieval, template listing/retrieval, performance audit, and connection info. The list/get pairs are conventional and the recipe-vs-template distinction is reinforced by descriptions.

    Naming Consistency4/5

    Most tools follow a clear verb_noun pattern (list_*, get_*) with plural list and singular get forms. Minor inconsistencies include audit_3d_web_performance mixing '3d' with the otherwise 'threejs' terminology, and connect_info being a noun-ish name rather than a strict verb_noun.

    Tool Count5/5

    Six tools is a well-scoped set for a specialized 3D recipe/template/performance resource. Each tool has a clear purpose and none feel redundant or missing at the surface level.

    Completeness4/5

    The server provides list/get coverage for both recipes and templates, plus an audit checklist and connection guidance, which covers the main discovery workflow. It lacks direct template source retrieval or search/filter capabilities, but those are not clearly required for this read-oriented MCP.

    Maintenance

    ActivitySlowing
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI agents to search for and fetch pre-vetted, production-safe 3D and motion components (React Three Fiber & GSAP) for injection into Next.js apps via AST-safe edits.
      MIT
    • F
      license
      Not graded
      quality
      D
      maintenance
      An AI-powered MCP server that generates production-ready React + Tailwind UI, 3D components, and full-stack integrations from natural language prompts inside Claude Desktop.
      -
    • A
      license
      A
      quality
      B
      maintenance
      Art-directed website prompts delivered as slash commands in Claude Code, Claude Desktop and Cursor: hero sections, landing pages, SaaS, portfolio, e-commerce and 3D/WebGL scenes. Each build is registered as a real MCP prompt and written natively for the tool you are in (Lovable, Bolt, Cursor or Claude), with a live demo you can scroll before spending a token.
      5
      58 npm
      1
      MIT
    • A
      license
      B
      quality
      C
      maintenance
      Enables creating immersive, interactive websites from scratch and transforming existing sites into immersive experiences, with scroll-driven frame scrubbing, business modules, a video/frame pipeline, and automatic headless verification. Works across any MCP client such as Claude Code, Cursor, VS Code, Gemini CLI, and Codex.
      27
      MIT