中文版 | 日本語版

An intelligent MCP (Model Context Protocol) server designed specifically for AI assistants to automatically log and manage conversation history with developers.

• 🤖 AI-Driven Logging - All content is determined and provided by the AI assistant
• 📝 Pure Save Mode - MCP only formats and stores, no content extraction or analysis
• 🔄 Designed for AI Retrospection - Log format optimized for AI to quickly understand project history
• 🏷️ Smart Organization - Auto-organize by project and date with tagging support
• 🔍 Powerful Search - Multi-dimensional search by keywords, files, tags, and time range
• 📊 Context Suggestions - Smart recommendations based on file associations

npm install

npm run build

Add MCP server configuration to Claude Code's config file (~/.claude.json):

{
  "mcpServers": {
    "conversation-logger": {
      "command": "node",
      "args": ["/path/to/ai-conversation-logger-mcp/dist/index.js"]
    }
  }
}

Restart Claude Code to apply the configuration.

Records every AI-user interaction with structured information:

interface LogConversationParams {
  userRequest: string;      // User's original request + uploaded file descriptions
  aiTodoList: string[];     // AI's execution plan (list even for view-only tasks)
  aiSummary: string;        // AI's operation summary (3-5 sentences)
  fileOperations?: string[]; // File operations in format: "action filepath - description"
  title?: string;           // Conversation title (optional)
  tags?: string[];          // Tag array (optional)
  project?: string;         // Project name (auto-detected if not provided)
}

Search through conversation history with multiple filters:

interface SearchParams {
  keywords?: string[];     // Keyword search
  filePattern?: string;    // File name pattern search
  days?: number;          // Recent N days
  project?: string;       // Project filter (defaults to current)
  tags?: string[];        // Tag filter
  limit?: number;         // Result limit (default: 10)
}

Get relevant historical context based on current work:

interface ContextParams {
  currentInput: string;    // Current user input
  currentFiles?: string[]; // Currently involved files
  project?: string;        // Project filter (optional)
}

Logs are stored in the project's ai-logs/ directory:

project-root/
├── ai-logs/
│   ├── 2025-08-07.md     # Daily conversation logs
│   ├── 2025-08-06.md
│   └── config.json       # Project configuration
├── src/
└── ...

Each conversation is recorded with the following structure:

## [Timestamp] Title #tags

### 🗣️ User Request
[Original user request]

### 📋 AI Execution Plan
- [x] Completed task
- [ ] Pending task

### 🤖 AI Summary
[Summary of what was accomplished]

### 📂 File Operations
- **Created** `path/to/file` - Purpose description
- **Modified** `path/to/file` - What was changed
- **Deleted** `path/to/file` - Reason for deletion

### 🏷️ Tags
#module #technology #type

All conversations should be logged, including:

• New feature development
• Bug fixes (any size)
• Code refactoring
• Configuration changes
• Code explanations and analysis
• Technical Q&A
• Code reviews
• Any project-related dialogue

1. AI-Driven Content - AI determines what information to log
2. Complete Context - Include all relevant details for future reference
3. Focus on "What" not "How" - Emphasize functionality over technical details
4. Consistent Format - Maintain standardized markdown structure

npm run dev

npm test

npm run lint
npm run lint:fix

npm run type-check

• TypeScript - Type-safe development
• MCP SDK - Model Context Protocol implementation
• Node.js - Runtime environment
• Jest - Testing framework

MIT

Contributions are welcome! Please feel free to submit a Pull Request.

For issues or suggestions, please open an issue on GitHub.