CAT
/MCP
SkillsMCPMarketplacesDigestToolsAdvertise

This week in Claude

Every Monday: Claude Code, Agent SDK, MCP, and the Anthropic platform moves worth your time.

Skills by Category
Frontend DevelopmentBackend & APIsTesting & QASecurityDevOps & CI/CDGit & Pull RequestsDocumentationCode Review & QualityAI & Agent BuildingSkill Development
MCP Servers by Category
Sales & MarketingWeb & Browser AutomationDatabasesAI & LLM ToolsCloud & InfrastructureCommunication & MessagingDeveloper ToolsDesign & CreativeDocuments & KnowledgeSearch & Web Crawling
Marketplaces by Category
AI Agents & OrchestrationLLM IntegrationDevelopment ToolsFrontend & UIBackend & APIsDatabasesTesting & Code QualityDevOps & CloudSecurity & ComplianceGit & Version Control

Cross AI Tools

Discover Claude Code plugins, extensions, and tools. Automatically updated directory of Anthropic Claude AI marketplaces with development tools, productivity plugins, and integrations.

Resources

  • Browse Skills
  • Browse MCP Servers
  • Browse Marketplaces
  • Plugins Reference

Community

  • About
  • Tools
  • Feedback
  • Privacy Policy
  • Advertise

Built for the Claude Code community with Claude Code by @mertduzgun

Independent project, not affiliated with Anthropic

Vue Harvest

virgilvox/vue-harvest
STDIOregistry active
Summary

Points at any Vue project and extracts a component library. Parses every .vue file with @vue/compiler-sfc to build a dependency graph, scores extraction confidence based on coupling patterns like direct store access or hardcoded APIs, and auto-extracts anything above your threshold into standalone bundles with rewritten imports. Also scrapes design tokens from all CSS and outputs them as custom properties plus a visual HTML explorer. The MCP server exposes twelve tools for the harder cases: deep_analyze for store-coupled components, suggest_refactor with before/after code, generate_wrapper for composable abstractions, and batch_triage to prioritize everything in the 30-70% confidence band. Reach for this when you're extracting a design system from an existing codebase or building an internal component registry.

CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
Keep your Mac awake
Keep your Mac awake
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Context.devContext.dev
Context.dev
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make money from your Skills
Make money from your Skills
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
AppSignal
AppSignal
Monitor with ease. Code with confidence.
Start Free Trial →
CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
Keep your Mac awake
Keep your Mac awake
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Context.devContext.dev
Context.dev
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make money from your Skills
Make money from your Skills
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
AppSignal
AppSignal
Monitor with ease. Code with confidence.
Start Free Trial →

vue-harvest

Point it at a Vue project. It finds every component, figures out which ones are reusable, and extracts them. It also pulls out your design tokens (colors, spacing, typography) and gives you a visual report.

The hard cases that need judgment (store-coupled components, ambiguous dependencies) get handled by an MCP server that feeds structured analysis to an LLM.

What it does

Component extraction. Parses every .vue file in a project using @vue/compiler-sfc. For each component it extracts the full interface (props, emits, slots), maps out the dependency graph, detects coupling issues, and scores a confidence level for safe extraction. Components above the threshold get auto-extracted into standalone bundles with rewritten imports and a manifest of peer dependencies.

Design token extraction. Scans all CSS (scoped styles, standalone stylesheets) and pulls out colors, font families, font sizes, font weights, spacing values, border radii, and shadows. Outputs a CSS custom properties file, a JSON dump, and a visual HTML explorer showing swatches and scales.

MCP server. Twelve tools that give an LLM the structured data it needs to reason about the harder cases: components in the 30-70% confidence range that could be extracted with some refactoring. The LLM can deep-analyze coupling, generate decoupling suggestions with before/after code, create composable wrappers for store-bound components, and do full rewrite-and-extract in one shot.

Install

npm install -g vue-harvest

# or run directly
npx vue-harvest analyze ./my-app

For the MCP server:

npm install -g vue-harvest-mcp

CLI

vue-harvest analyze [path]

Full analysis pipeline. Discovers .vue files, parses interfaces, builds the dependency graph, classifies every component, auto-extracts the safe ones, and generates a registry + catalog.

vue-harvest analyze
vue-harvest analyze ./path/to/project
vue-harvest analyze --threshold 60
vue-harvest analyze --output ./harvested
vue-harvest analyze --json | jq '.summary'

Output goes to .vue-harvest/ by default:

.vue-harvest/
  registry.json       Component registry (shadcn-compatible format)
  catalog.html        Browsable catalog with tier filters
  analysis.json       Full analysis dump for the MCP server
  SUMMARY.md          Human-readable report
  components/         Extracted component bundles
    Button/
      Button.vue
      manifest.json
    Card/
      Card.vue
      manifest.json

vue-harvest list [path]

Lists all components sorted by extraction confidence.

vue-harvest list
vue-harvest list --tier primitive
vue-harvest list --json

vue-harvest inspect <name>

Full breakdown of a single component: props, events, slots, dependencies, coupling issues, style analysis.

vue-harvest inspect Button
vue-harvest inspect UserProfileCard --json

vue-harvest extract <name>

Extracts a specific component and all its local dependencies into a standalone bundle.

vue-harvest extract Button
vue-harvest extract UserProfileCard --force
vue-harvest extract Card --output ./my-components

vue-harvest tokens [path]

Extracts design system tokens from the project.

vue-harvest tokens
vue-harvest tokens --json
vue-harvest tokens --output ./design-system

Outputs:

  • tokens.css with CSS custom properties
  • tokens.json with the full token dataset
  • design-system.html with a visual explorer (color swatches, font scales, spacing visualization)

vue-harvest init

Creates a harvest.config.json with detected project settings.

Configuration

Optional. Place a harvest.config.json in the project root:

{
  "include": ["**/*.vue"],
  "exclude": [
    "node_modules/**",
    "dist/**",
    "**/*.test.*",
    "**/*.story.*"
  ],
  "extractionThreshold": 70,
  "outDir": ".vue-harvest",
  "registry": "json"
}

Path aliases are auto-detected from tsconfig.json. If your project has @ mapped to ./src, vue-harvest picks that up.

Component classification

Every component gets classified into a reusability tier based on its interface, dependencies, and coupling:

TierConfidenceWhat it means
Primitive85-100%Pure UI, no business logic. Button, Input, Card.
Composite70-85%Built from primitives, minimal logic. FormField, DataTable.
Feature50-70%Has business logic but potentially reusable. UserAvatar, SearchBar.
Page-bound30-50%Tightly coupled to a specific page or route.
App-specific0-30%Deeply coupled to app state. Not reusable as-is.

Components at or above the extraction threshold (default 70%) are auto-extracted. The 30-70% band is where the MCP server comes in.

Coupling issues

The analyzer detects these coupling patterns:

IssueSeverityDescription
direct-store-accesswarningImports a Pinia/Vuex store directly
hardcoded-apiwarningContains hardcoded API endpoint URLs
router-dependencyinfoUses vue-router
i18n-dependencywarningUses vue-i18n
global-injectwarningUses inject() for app-level provides
env-variablewarningReferences import.meta.env
unscoped-csswarningHas unscoped styles that leak globally
deep-provide-chainwarningRelies on provide/inject chains
implicit-globalwarningUses globally registered components without importing them
side-effect-importwarningHas imports that execute side effects

MCP server

Setup with Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "vue-harvest": {
      "command": "vue-harvest-mcp"
    }
  }
}

Setup with Claude Code

claude mcp add vue-harvest vue-harvest-mcp

Available tools

ToolWhat it does
analyze-projectRuns the full pipeline on a project path
list-componentsLists components with filters (tier, confidence range)
inspect-componentFull analysis with source code
extract-componentExtracts with force option
deep-analyzeStructured coupling analysis for LLM reasoning
suggest-refactorBefore/after code for decoupling
generate-wrapperCreates composable wrappers for store-bound components
adapt-and-extractFull rewrite + extract in one step
batch-triagePrioritizes all reviewable components with effort estimates
coupling-reportProject-wide coupling patterns
analyze-design-systemExtracts design tokens
get-design-tokensReturns tokens filtered by type

Resources

The server exposes these as MCP resources after analysis:

  • harvest://registry : full registry JSON
  • harvest://graph : dependency graph
  • harvest://summary : analysis summary
  • harvest://component/{name} : individual component analysis
  • harvest://design-system : extracted design tokens

Prompts

  • analyze-new-project : guided first analysis
  • extraction-sprint : batch refactor and extract session
  • refactor-component : single component deep refactor
  • extract-design-system : design token extraction and analysis

Example conversations

"Analyze my Vue project at /Users/me/projects/my-app"

"Show me all the components that need review"

"Deep analyze the UserProfileCard component and suggest how to decouple it from the auth store"

"Do an extraction sprint, go through all reviewable components and extract what you can"

"Extract the design system tokens and recommend a naming convention"

Programmatic API

vue-harvest exports its analysis engine for use in other tools:

import { analyze, writeOutput, analyzeTokens } from 'vue-harvest'

const report = await analyze('./my-vue-app', {
  extractionThreshold: 0.6,
})

console.log(report.summary)
// { totalFiles: 42, analyzed: 40, autoExtracted: 12, needsMCP: 8, ... }

await writeOutput(report)

// Design tokens
const tokens = await analyzeTokens('./my-vue-app')
console.log(tokens.palette)    // [{ hex: '#3b82f6', usageCount: 14 }, ...]
console.log(tokens.spacing)    // ['4px', '8px', '12px', '16px', '24px']

Architecture

vue-harvest (monorepo)
  packages/
    cli/                        Published as "vue-harvest" on npm
      src/
        types.ts                Type system shared across the whole project
        index.ts                Pipeline orchestrator
        cli.ts                  CLI entry point (citty)
        analyzers/
          sfc-analyzer.ts       SFC parsing, interface extraction, coupling detection
          graph-builder.ts      Dependency graph with Tarjan's SCC for cycle detection
          design-system-analyzer.ts   Token extraction from CSS
        extractors/
          component-extractor.ts   Import rewriting, file bundling, manifest generation
        generators/
          registry.ts           Registry JSON + catalog HTML generation
        commands/               CLI command definitions
        utils/
          config.ts             Config resolution, alias detection from tsconfig
      tests/
        fixtures/               A real mini Vue project used as test input
        unit/                   66 tests across 5 suites

    mcp/                        Published as "vue-harvest-mcp" on npm
      src/
        index.ts                MCP server (12 tools, resources, prompts)

The split between CLI and MCP is deliberate. The CLI handles everything deterministic: parsing, graph building, classification, extraction. The MCP handles the 20% that needs reasoning: decoupling suggestions, refactoring code generation, ambiguity resolution.

Development

git clone https://github.com/virgilvox/vue-harvest.git
cd vue-harvest
pnpm install
pnpm build
pnpm test

Watch mode for the CLI:

pnpm --filter vue-harvest dev

Run a specific test file:

npx vitest run packages/cli/tests/unit/sfc-analyzer.test.ts

How it works internally

  1. Discovery. Globs for .vue files, reads tsconfig.json for path aliases.
  2. SFC parsing. @vue/compiler-sfc splits each file into template, script, and style blocks.
  3. Interface extraction. Regex-based extraction of defineProps, defineEmits, <slot> tags. Handles generic type syntax, withDefaults, and object syntax with nested options.
  4. Dependency analysis. es-module-lexer parses imports. Each import gets classified by kind (internal component, composable, store, util, external package, etc).
  5. Coupling detection. Pattern matching on the script and template AST for store access, hardcoded APIs, router usage, i18n, inject, env vars, unscoped CSS.
  6. Classification. Confidence scoring starts at 1.0 and gets penalized for coupling issues, store access, and page-path patterns. Boosted for props/slots interfaces and UI-path patterns.
  7. Graph building. Resolves import paths, builds adjacency lists, runs Tarjan's strongly connected components algorithm for cycle detection, computes transitive dependency closures.
  8. Extraction. For components above threshold: collects the component plus all its local deps (composables, utils, types, styles), rewrites import paths to be relative within the extracted bundle, generates a manifest listing peer dependencies and required globals.
  9. Token extraction. PostCSS value parser identifies colors (hex, rgb, hsl), font properties, spacing values, radii, and shadows from CSS declarations. Deduplicates, normalizes hex values, and groups by type.

License

MIT

Featured
CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
Keep your Mac awake
Keep your Mac awake
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Context.devContext.dev
Context.dev
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make money from your Skills
Make money from your Skills
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
AppSignal
AppSignal
Monitor with ease. Code with confidence.
Start Free Trial →
Categories
Design & CreativeSearch & Web Crawling
Registryactive
Packagevue-harvest-mcp
TransportSTDIO
UpdatedMar 31, 2026
View on GitHub

Related Design & Creative MCP Servers

View all →
HTML to Figma — Design System

miapre/html-to-figma-design-system

Translate HTML prototypes into Figma using your design system's real components and tokens.
3
Illustrator Mcp Server

ie3jp/illustrator-mcp-server

Read, manipulate, and export Adobe Illustrator design data. 26 tools. macOS | Windows.
44
Godot

coding-solo/godot-mcp

MCP server for interfacing with Godot game engine. Provides tools for launching the editor, running projects, and capturing debug output.
3.7k
Unity Mcp

ivanmurzak/unity-mcp

Make 3D games in Unity Engine with AI. MCP Server + Plugin for Unity Editor and Unity games.
3.1k
Excalidraw

yctimlin/mcp_excalidraw

Provides an Excalidraw canvas exposed via MCP for real-time diagramming and element CRUD from AI agents.
1.9k
Figma MCP Server

figma/mcp-server-guide

The Figma MCP server brings Figma design context directly into your AI workflow.
1.6k