eds-mcp-server

eds-mcp-server

An MCP server that supercharges AEM Edge Delivery Services development in IDEs by providing tools for scaffolding blocks, validating code, and generating configurations directly through natural language.

Category
访问服务器

README

AEM Edge Delivery Services — MCP Server

A Model Context Protocol server that supercharges AEM EDS development in VS Code, Cursor, and any MCP-compatible IDE. No API keys needed — your IDE's own LLM uses these tools, resources, and prompts to generate EDS-compliant code.

What It Does

Category Tools Description
Scaffolding scaffold_block Generate complete block files (JS, CSS, README, test.html, sample content)
scaffold_model Generate Universal Editor component model/definition/filter JSON
scaffold_project Step-by-step guide for new EDS projects (standard & repoless)
generate_block_from_design Multimodal — turn a text description, design image, and/or Figma URL into an EDS block using Adobe's Content-Driven-Development workflow
Validation validate_block Check JS, CSS, JSON model, and content against EDS standards
check_performance Analyze block code for performance issues and budget impact
Guidance explain_dom Show how authored content tables transform into DOM
lookup_block Search block patterns and find existing implementations
search_block_collection Search Adobe Block Collection & Block Party via GitHub API
eds_config Generate configuration files (fstab, redirects, headers, etc.)
eds_scripts_guide Guidance for scripts.js, delayed.js, and lifecycle customization

Plus 4 resources (coding standards, block guide, cheatsheet, Adobe EDS skills) and 3 prompt templates (new-block, fix-block, design-to-block).

Prompt Parameters

Prompt Parameters
new-block blockName (required), description (required)
fix-block blockName (required), issue (required)
design-to-block blockName (required), text, imageRefs, figmaUrl (at least one of the last three)

Design → Block workflow

The generate_block_from_design tool and the design-to-block prompt integrate Adobe's official EDS skills (Content-Driven Development, analyze-and-plan, content-modeling, authoring-analysis, building-blocks, UE component model, testing-blocks, code-review).

Provide any combination of:

  • Text — natural-language description of what the block should do
  • Image(s) — local paths or URLs of design screenshots/mockups (the IDE LLM analyzes them with vision)
  • Figma URL — a figma.com/file/... or figma.com/design/... link, optionally with ?node-id=…. The tool emits a local curl recipe to pull node JSON + a 2× PNG export; your Figma token stays on your machine.

The tool returns: the CDD workflow outline, a vision-analysis prompt for the IDE LLM, the Figma-fetch recipe (if applicable), implementation patterns, a baseline scaffold (JS/CSS/README/test.html/sample-content), the UE component-model guidance, a testing matrix, and a self-review checklist.

Quick Start

Cursor

Open Settings → Tools & Integrations → MCP Servers → Add and paste:

{
  "mcpServers": {
    "eds-dev": {
      "command": "npx",
      "args": ["-y", "@anthropic-eds/eds-mcp-server"]
    }
  }
}

Or create .cursor/mcp.json in your project root with the same config.

VS Code

Create .vscode/mcp.json in your project:

{
  "servers": {
    "eds-dev": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@anthropic-eds/eds-mcp-server"]
    }
  }
}

Then use GitHub Copilot Chat in Agent Mode (the @ button) to access the tools.

Local CLI Validator (New)

This project now includes an EDS block validator/linter CLI.

npm run build
npm run validate:blocks -- ./blocks
# or directly
node dist/cli.js ./blocks --strict

Options:

  • --json for CI-friendly machine output
  • --strict to fail on warnings
  • --help for usage details

Claude Code

Add to your ~/.claude/mcp.json:

{
  "mcpServers": {
    "eds-dev": {
      "command": "npx",
      "args": ["-y", "@anthropic-eds/eds-mcp-server"]
    }
  }
}

From Source (Development)

git clone https://github.com/your-org/eds-mcp-server.git
cd eds-mcp-server
npm install
npm run build

# Test with MCP Inspector
npm run inspect
# Opens browser at http://localhost:6274

# Add to IDE pointing to local build:
# "command": "node",
# "args": ["/path/to/eds-mcp-server/dist/index.js"]

Usage Examples

Scaffold a new block

Ask your IDE's AI: "Create a hero block with a background image, heading, and CTA button"

The LLM will call scaffold_block and return all 5 files ready to copy into your project.

Validate existing code

Ask: "Validate my cards block" — paste your JS and CSS. The validate_block tool checks for scoping issues, missing exports, reserved class names, performance problems, and more.

Understand the DOM pipeline

Ask: "How does a 3-row, 2-column table become DOM in EDS?"

The explain_dom tool shows the exact HTML structure your decorate function receives.

Set up a new project

Ask: "Set up a new repoless EDS site with Google Drive and Universal Editor"

The scaffold_project tool returns a complete step-by-step guide.

Get configuration templates

Ask: "I need to set up redirects and custom headers"

The eds_config tool returns ready-to-use spreadsheet formats and configuration examples.

Check performance

Ask: "Is this block going to hurt my Lighthouse score?" — paste the code.

The check_performance tool estimates budget impact and flags render-blocking patterns.

Tool Reference

scaffold_block

Parameter Type Required Default Description
blockName string Block name in kebab-case
description string What the block does
variant string Variant name (e.g. "dark")
layout grid|flex|stack stack CSS layout strategy
hasMedia boolean false Has image/video column
interactive boolean false Needs event handlers
fields array Universal Editor field definitions (name, type, label)

scaffold_model

Parameter Type Required Description
blockName string Block name in kebab-case
title string Human-readable block title for the editor
group string Block group/category for editor UI organization
fields array Block fields (name, type, label, required)
allowedChildren string[] Component IDs allowed as children (for container blocks)

scaffold_project

Parameter Type Required Default Description
projectType enum standard, repoless-code, repoless-content
siteName string Project/site name
contentSource enum google-drive google-drive or sharepoint
includeUE boolean false Include Universal Editor component model files
includeCommerce boolean false Include Commerce Drop-in integration setup

validate_block

Parameter Type Required Description
blockName string Block name
js string Block JS contents
css string Block CSS contents
json string Component model JSON
content string Sample content markdown

check_performance

Parameter Type Required Default Description
blockName string Block name
js string Block JS file contents
css string Block CSS file contents
isAboveFold boolean false Whether block appears above the fold (eager-loaded)

explain_dom

Parameter Type Required Default Description
blockName string Block name
variant string Block variant (e.g. "dark", "wide")
rows number 2 Number of content rows (1–20)
columns number 2 Number of columns per row (1–6)
cellContents string[][] Specific cell contents as 2D array [row][col]

lookup_block

Parameter Type Required Description
query string Block name or description to search for (e.g. "hero", "tabbed content")

search_block_collection

Parameter Type Required Default Description
query string Block name or type to search for
source enum both collection (Adobe official), party (community), or both

eds_config

Parameter Type Required Default Description
configType enum fstab, fstab-sharepoint, fstab-repoless, head-html, redirects, headers, robots, sitemap, helix-config, metadata, repoless, all
domain string Your site domain
contentSource enum google-drive google-drive or sharepoint
folderId string Google Drive folder ID or SharePoint path

eds_scripts_guide

Parameter Type Required Description
topic enum scripts-js-overview, load-eager, load-lazy, load-delayed, auto-blocking, metadata, header-footer, custom-fonts, analytics, third-party, decorateMain, all

Architecture

eds-mcp-server/
├── src/
│   ├── index.ts              # Server entry point — registers all tools
│   ├── tools/
│   │   ├── scaffold-block.ts   # Generate block files
│   │   ├── scaffold-model.ts   # Generate UE model files
│   │   ├── scaffold-project.ts # New project guide
│   │   ├── validate-block.ts   # Block validation
│   │   ├── explain-dom.ts      # Content → DOM mapping
│   │   ├── eds-config.ts       # Configuration templates
│   │   ├── check-performance.ts # Performance analysis
│   │   ├── lookup-block.ts     # Block pattern search
│   │   ├── search-block-collection.ts # GitHub API block search
│   │   └── eds-scripts.ts      # Scripts customization guide
│   ├── resources/
│   │   └── eds-resources.ts    # Documentation as MCP resources
│   ├── prompts/
│   │   └── eds-prompts.ts      # Prompt templates
│   └── knowledge/
│       ├── eds-conventions.ts  # EDS rules, constraints, templates
│       └── block-templates.ts  # Code generators
├── package.json
├── tsconfig.json
├── LICENSE
├── evaluations.xml
└── README.md

Key design decisions:

  • No LLM dependency — the server provides tools and knowledge; your IDE's LLM does the reasoning
  • No API keys — pure static analysis, scaffolding, and documentation (except search_block_collection which queries public GitHub API)
  • Convention-encoded — EDS rules are hardcoded into validators and generators
  • Tool annotations — all tools declare readOnlyHint, destructiveHint, idempotentHint, and openWorldHint
  • Stdio transport — works with any MCP client (Cursor, VS Code, Claude Code)

Evaluations

The evaluations.xml file contains 10 LLM-facing integration test Q&A pairs. Each question requires calling one or more MCP tools and verifying the response matches expected behavior. Use these with the MCP Inspector to validate the server end-to-end.

EDS Resources

Contributing

  1. Fork the repository
  2. Create a feature branch
  3. Add tools in src/tools/, knowledge in src/knowledge/
  4. Test with npm run inspect
  5. Submit a PR

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选