Vibe Coding Documentation MCP (MUSE)
Automatically collects, summarizes, and documents code from vibe coding sessions, generating multiple document types (README, DESIGN, TUTORIAL, etc.) and publishing them to platforms like Notion, GitHub Wiki, and Obsidian.
README
Vibe Coding Documentation MCP (MUSE)
MCP server that automatically collects, summarizes, documents, and publishes code and design decisions created during vibe coding sessions.
Features
This MCP server provides 15 tools for managing vibe coding documentation:
| Tool | Description |
|---|---|
muse_collect_code_context |
Collects code blocks and conversation summaries into structured context |
muse_summarize_design_decisions |
Extracts key architectural and design decisions from conversation logs |
muse_generate_dev_document |
Generates README, DESIGN, TUTORIAL, CHANGELOG, API, or ARCHITECTURE documents |
muse_normalize_for_platform |
Converts Markdown documents for Notion, GitHub Wiki, or Obsidian |
muse_publish_document |
Publishes generated documents to external platforms |
muse_create_session_log |
Creates daily or session-based vibe coding session logs |
muse_analyze_code |
AST-based code analysis with Mermaid diagram generation |
muse_session_history |
Manages session history - save, retrieve, search past sessions |
muse_export_session |
Exports sessions to Markdown, JSON, or HTML formats |
muse_project_profile |
Manages project-specific settings and configurations |
muse_git |
Git integration - status, log, diff, branch, snapshot, design decision extraction |
muse_session_stats |
Session analytics dashboard with productivity insights and trends |
muse_auto_tag |
AI-powered auto-tagging for sessions based on content analysis |
muse_template |
Custom template management for documents and reports |
muse_batch |
Batch operations to execute multiple tools in sequence or parallel |
Additional Features (v2.0)
- AST Parsing: TypeScript, Python, Go code analysis
- Mermaid Diagrams: Class, Flowchart, Sequence, ER, Architecture diagrams
- Multi-language: Korean/English support
- 6 Document Types: README, DESIGN, TUTORIAL, CHANGELOG, API, ARCHITECTURE
- 6 Platforms: Notion, GitHub Wiki, Obsidian, Confluence, Slack, Discord
Code Quality (v2.1)
- Input Validation: Zod schema-based type-safe validation for all tools
- Error Handling: Structured error classes (ToolError, ValidationError, PlatformError)
- Security: Command injection prevention (exec → spawn), path sanitization
- Performance: LRU cache, regex cache, memoization utilities
Security (v2.2)
- Path Traversal Prevention: Validates file paths stay within allowed directories
- SSRF Protection: Webhook URL validation for Slack/Discord
- Network Timeout: AbortController-based request timeout (30s default)
- Retry Logic: Exponential backoff with configurable retry attempts
Enhanced Quality (v2.3)
- Structured Logging: JSON-based logging with child loggers per tool
- Configuration Validation: Startup validation for all platform configurations
- Platform Expansion: Full support for 6 platforms (Notion, GitHub Wiki, Obsidian, Confluence, Slack, Discord)
- AST Memoization: Cached code analysis for improved performance
- Test Coverage: 149 tests with 85%+ coverage on core modules
AI-Powered Analysis (v2.4)
- Claude AI Integration: Use Claude AI for enhanced design decision analysis
- Smart Summarization: AI-generated insights and recommendations
- Fallback Support: Automatic fallback to pattern-based analysis when AI unavailable
- Optional Feature: Enable with
useAI: trueparameter
AI Code Analysis (v2.5)
- AI-Powered Code Review: Deep code analysis with quality, security, and performance insights
- Issue Detection: Identify potential bugs, security vulnerabilities, and code smells
- Improvement Suggestions: AI-generated recommendations for better code
- Works with AST: Combines AI insights with AST-based analysis for comprehensive results
Session History (v2.6)
- Persistent Storage: Save coding sessions to local JSON files
- CRUD Operations: Create, read, update, delete sessions
- Search & Filter: Find past sessions by keyword, tags, or date
- Statistics: Track total sessions, code contexts, and design decisions
Session Export & Project Profiles (v2.7)
- Session Export: Export sessions to Markdown, JSON, or HTML formats
- Multiple Templates: Default, minimal, detailed, and report templates
- Project Profiles: Manage project-specific settings and configurations
- Publishing Config: Default platforms and settings per project
- Code Analysis Config: Language preferences and diagram types
- Documentation Config: Default document types, language, author info
- Team Management: Store team member information per project
Git Integration (v2.8)
- Repository Status: View staged, unstaged, and untracked files
- Commit History: Browse commit log with filtering by author, date, grep
- Diff Analysis: View changes with statistics (staged, unstaged, between refs)
- Branch Info: List local and remote branches with tracking info
- Git Snapshot: Capture complete repository state for session context
- Design Decision Extraction: Auto-extract design decisions from commit messages
- Session Linking: Attach Git context to coding sessions
Session Statistics Dashboard (v2.9)
- Overview Analytics: Total sessions, code contexts, design decisions at a glance
- Language Distribution: Breakdown of programming languages used across sessions
- Timeline View: Session activity over time (daily, weekly, monthly)
- Tag Analytics: Most used tags and tag co-occurrence analysis
- Productivity Insights: Session duration, code output, and efficiency metrics
- Trend Analysis: Compare current period with previous or average
AI Auto-tagging (v2.10)
- Smart Tag Suggestions: Pattern-based and AI-powered tag recommendations
- Confidence Scoring: Each tag suggestion includes confidence level
- Custom Tag Training: Train the system with custom tag patterns
- Configurable Rules: Define tag rules for file extensions, keywords, patterns
- Batch Tagging: Apply tags to multiple sessions at once
Custom Templates (v2.11)
- Template Management: Create, edit, delete custom document templates
- Variable Substitution: Support for
{{variable}},${variable},{variable}formats - Built-in Templates: README, Session Summary, Weekly Report templates included
- Template Preview: Preview rendered output before applying
- Import/Export: Share templates between projects
Batch Operations (v2.12)
- Sequential Execution: Run multiple tools in order with dependency management
- Parallel Execution: Execute independent operations concurrently
- Dependency Graph: Topological sort for operation ordering
- Job Tracking: Monitor batch job status, cancel running jobs
- Error Handling: Stop on error or continue with remaining operations
- Result Chaining: Pass output from one operation to next using
$refsyntax
Installation
Claude Code (Recommended)
claude mcp add vibe-coding-mcp npx vibe-coding-mcp
npm
npm install -g vibe-coding-mcp
Claude Desktop
Add to claude_desktop_config.json:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"vibe-coding-mcp": {
"command": "npx",
"args": ["vibe-coding-mcp"]
}
}
}
Environment Variables
# Anthropic API (optional, for AI-powered analysis)
ANTHROPIC_API_KEY=your_anthropic_api_key_here
# Notion API (optional)
NOTION_API_KEY=your_notion_api_key_here
NOTION_DATABASE_ID=your_database_id_here
# GitHub (optional, for Wiki publishing)
GITHUB_TOKEN=your_github_token_here
GITHUB_REPO=owner/repo
# Confluence (optional)
CONFLUENCE_BASE_URL=https://your-domain.atlassian.net
CONFLUENCE_USERNAME=your_email@example.com
CONFLUENCE_API_TOKEN=your_api_token_here
CONFLUENCE_SPACE_KEY=YOURSPACE
# Slack (optional, webhook URL passed via tool parameter)
SLACK_WEBHOOK_URL=https://hooks.slack.com/services/...
# Discord (optional, webhook URL passed via tool parameter)
DISCORD_WEBHOOK_URL=https://discord.com/api/webhooks/...
Demo Scenarios
1. Generate README and Publish to Notion
User: Collect the code we wrote today and create a README, then publish to Notion.
Claude: [Uses collect_code_context → generate_dev_document → normalize_for_platform → publish_document]
2. Create Design Decision Docs for GitHub Wiki
User: Summarize our design decisions and publish to GitHub Wiki.
Claude: [Uses summarize_design_decisions → generate_dev_document → normalize_for_platform → publish_document]
3. Daily Vibe Coding Log
User: Create a session log for today's work.
Claude: [Uses collect_code_context → create_session_log]
4. Git-Aware Session Documentation
User: Capture my current Git state and create a session with design decisions from commits.
Claude: [Uses muse_git(action='snapshot') → muse_git(action='extractDecisions') → muse_session_history(action='save')]
5. Complete Session Export with Git Context
User: Export all my sessions from this week with Git information.
Claude: [Uses muse_git(action='linkToSession') → muse_export_session(format='markdown')]
6. Session Analytics Dashboard
User: Show me my coding productivity statistics for this month.
Claude: [Uses muse_session_stats(action='overview') → muse_session_stats(action='productivity') → muse_session_stats(action='trends')]
7. Auto-tag and Organize Sessions
User: Analyze my sessions and suggest relevant tags.
Claude: [Uses muse_auto_tag(action='suggest') → muse_auto_tag(action='apply')]
8. Generate Custom Report with Template
User: Create a weekly report using the team report template.
Claude: [Uses muse_template(action='apply', templateId='weekly-report') → muse_publish_document]
9. Batch Documentation Workflow
User: Analyze this code, generate docs, and publish to Notion in one go.
Claude: [Uses muse_batch(action='execute', operations=[
{tool: 'muse_analyze_code', params: {...}},
{tool: 'muse_generate_dev_document', params: {...}, dependsOn: ['op_0']},
{tool: 'muse_publish_document', params: {...}, dependsOn: ['op_1']}
])]
Supported Platforms
- Notion: Full API integration with page creation
- GitHub Wiki: Git-based wiki updates
- Obsidian: Local vault file storage with frontmatter support
- Confluence: Atlassian Confluence page publishing
- Slack: Webhook-based message publishing
- Discord: Webhook-based message publishing
Project Structure
src/
├── stdio.ts # MCP server entry point (stdio transport)
├── index.ts # HTTP/SSE server entry point
├── core/
│ ├── schemas.ts # Zod validation schemas
│ ├── errors.ts # Structured error classes
│ ├── cache.ts # LRU cache & memoization
│ ├── security.ts # Path traversal, SSRF, timeout utilities
│ ├── logger.ts # Structured JSON logging
│ └── config.ts # Platform configuration validation
├── tools/ # 15 MCP tools
├── platforms/ # Notion, GitHub Wiki, Obsidian, Confluence, Slack, Discord
├── types/ # TypeScript interfaces
└── utils/
├── markdown.ts # Markdown processing
├── astParser.ts # AST parsing for TypeScript, Python, Go
├── diagramGenerator.ts # Mermaid diagram generation
├── gitExecutor.ts # Safe Git command execution
└── gitParsers.ts # Git output parsing utilities
Development
# Watch mode
npm run dev
# Build
npm run build
# Start (HTTP/SSE mode)
npm start
# Start (stdio mode for Claude Desktop)
npm run stdio
# Run tests
npm test
# Run tests with coverage
npm run test:coverage
Dependencies
| Package | Purpose |
|---|---|
@modelcontextprotocol/sdk |
MCP server SDK |
@notionhq/client |
Notion API integration |
zod |
Input validation |
typescript |
TypeScript compiler |
License
MIT
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。