Agent MCP

# Agent-MCP Pre-Launch Configuration TUI This document describes the interactive pre-launch TUI that asks you to configure your tools and settings **before** the server starts, giving you complete control over which features to enable. ## How It Works 1. **Run the server**: `npm run server` 2. **TUI appears** before server starts 3. **Choose your configuration** with interactive menus 4. **Server starts** with exactly what you selected ## Configuration Options The TUI provides multiple ways to configure your setup: ### 🚀 **Quick Start** - One-click setup with recommended Memory + RAG mode - Perfect for most users who want code assistance - Lightweight and fast startup ### 🎯 **Predefined Modes** - Choose from optimized configurations - See tool counts and performance estimates - Full/Memory+RAG/Minimal/Development modes ### 🔧 **Custom Configuration** - Pick exactly which tool categories you want - Granular control over functionality - See descriptions for each category ### 📊 **Advanced Setup** - Detailed walkthrough with explanations - Category-by-category configuration - Shows dependencies and warnings ## Predefined Modes ### Full Mode (Default) - **Description**: Complete agent orchestration platform with all features - **Tools**: 33 tools across 9 categories - **Use Case**: Full multi-agent development workflows - **Categories**: All enabled ### Memory + RAG Mode - **Description**: Lightweight mode focused on knowledge and memory management - **Tools**: 15 tools across 5 categories - **Use Case**: Code assistance with contextual memory but no agent orchestration - **Categories**: basic, rag, memory, fileManagement, sessionState ### Minimal Mode - **Description**: Essential tools only for basic functionality - **Tools**: 1 tool (health check only) - **Use Case**: Testing, debugging, or as a base for custom configurations - **Categories**: basic only ### Development Mode - **Description**: Development-focused tools without agent management complexity - **Tools**: Varies, typically 18-20 tools - **Use Case**: Solo development with RAG and memory but no multi-agent features - **Categories**: basic, rag, memory, fileManagement, sessionState, assistanceRequest ## Tool Categories | Category | Description | Tools | |----------|-------------|--------| | `basic` | Health checks, system status (always enabled) | 1 | | `rag` | Vector search, knowledge base, retrieval augmented generation | 2 | | `memory` | Project context storage, persistent memory | 6 | | `agentManagement` | Agent creation, termination, lifecycle management | 7 | | `taskManagement` | Task creation, assignment, workflows | 6 | | `fileManagement` | File operations, content access | 2 | | `agentCommunication` | Inter-agent messaging, collaboration | 3 | | `sessionState` | Session persistence, recovery, state management | 4 | | `assistanceRequest` | Intelligent assistance routing | 1 | ## CLI Usage ### Interactive Configuration ```bash # Launch TUI configuration interface npm run server --config-mode # Or via built binary agent-mcp --config-mode ``` ### Predefined Modes ```bash # Use specific mode npm run server --mode memoryRag npm run server --mode minimal npm run server --mode development npm run server --mode full # Skip TUI and use saved/default config npm run server --no-tui ``` ### Environment Variables ```bash # Skip TUI entirely AGENT_MCP_SKIP_TUI=true npm run server # Override specific categories AGENT_MCP_ENABLE_RAG=false npm run server AGENT_MCP_ENABLE_AGENTS=true npm run server # CI/CD environments automatically skip TUI CI=true npm run server ``` ## Configuration File Configuration is automatically saved to `.agent/tool-config.json`: ```json { "version": "1.0", "mode": "memoryRag", "categories": { "basic": true, "rag": true, "memory": true, "agentManagement": false, "taskManagement": false, "fileManagement": true, "agentCommunication": false, "sessionState": true, "assistanceRequest": false }, "lastModified": "2025-01-09T17:30:00.000Z" } ``` ## HTTP API Endpoints ### Health Check with Configuration ```bash GET /health ``` Response includes configuration information: ```json { "status": "healthy", "configuration": { "mode": "memoryRag", "enabled_categories": ["basic", "rag", "memory", "fileManagement", "sessionState"], "total_categories": 9 }, "tools": ["health", "ask_project_rag", "get_rag_status", "..."], "...": "other health data" } ``` ### Detailed Statistics ```bash GET /stats ``` Response includes comprehensive configuration details: ```json { "configuration": { "mode": "memoryRag", "enabled_categories": ["basic", "rag", "memory", "fileManagement", "sessionState"], "disabled_categories": ["agentManagement", "taskManagement", "agentCommunication", "assistanceRequest"], "total_categories": 9, "details": { "basic": true, "rag": true, "memory": true, "agentManagement": false, "...": "full category details" } }, "tools": { "total_tools": 15, "tool_names": ["health", "ask_project_rag", "..."] }, "...": "other system stats" } ``` ## TUI Interface Features The interactive TUI provides: 1. **Current Configuration Review** - Shows active mode and enabled categories 2. **Predefined Mode Selection** - Quick setup with common configurations 3. **Custom Configuration** - Individual category selection with descriptions 4. **Configuration Validation** - Warnings for potentially problematic setups 5. **Impact Assessment** - Memory usage and startup time estimates ### TUI Navigation - Use arrow keys to navigate options - Space bar to select/deselect in checkbox lists - Enter to confirm selections - The interface provides helpful descriptions for each tool category ## Benefits ### Performance - **Reduced Memory Usage**: Only load required tools (15 tools vs 33 in minimal setups) - **Faster Startup**: Fewer imports and initializations (2-6s vs full 6s) - **Lower Resource Consumption**: Ideal for constrained environments ### Flexibility - **Use Case Optimization**: Match configuration to specific needs - **Gradual Adoption**: Start minimal and add features as needed - **Environment Adaptation**: Different configs for dev/staging/production ### Usability - **Interactive Setup**: No need to memorize configuration options - **Clear Documentation**: Built-in help and descriptions - **Persistent Settings**: Configuration saved between sessions ## Migration from Previous Versions Existing installations will: 1. Continue working with full mode (all tools enabled) 2. Show TUI on first run if no configuration exists 3. Can opt-out of TUI with `--no-tui` or `AGENT_MCP_SKIP_TUI=true` ## Advanced Usage ### Custom Mode Creation While not directly exposed in the TUI, you can manually edit `.agent/tool-config.json` to create custom configurations that don't match any predefined mode. ### Integration with CI/CD ```yaml # GitHub Actions example env: CI: true AGENT_MCP_SKIP_TUI: true run: | npm run server --mode minimal --no-tui ``` ### Docker Deployment ```dockerfile ENV AGENT_MCP_SKIP_TUI=true ENV AGENT_MCP_ENABLE_RAG=true ENV AGENT_MCP_ENABLE_MEMORY=true CMD ["npm", "run", "server", "--no-tui"] ``` ## Troubleshooting ### Configuration Issues - Use `--mode full` to reset to default configuration - Delete `.agent/tool-config.json` to force TUI on next startup - Check `/health` endpoint to verify current configuration ### TUI Not Showing - Ensure you're not in CI environment (`CI` env var) - Check that `AGENT_MCP_SKIP_TUI` is not set to `true` - Use `--config-mode` to force TUI launch ### Permission Issues - Ensure write access to `.agent/` directory - Configuration requires ability to create/modify files in project directory ## Implementation Details The system uses: - **Dynamic imports** for conditional tool loading - **Zod schemas** for configuration validation - **Commander.js** for CLI argument parsing - **Inquirer.js** for interactive TUI interface - **JSON persistence** for configuration storage This enables flexible, performant, and user-friendly tool configuration management.