Enhanced Architecture MCP
Enhanced Model Context Protocol (MCP) servers with professional accuracy, tool safety, user preferences, and intelligent context monitoring.
Overview
This repository contains a collection of MCP servers that provide advanced architecture capabilities for AI assistants, including:
- Professional Accuracy Enforcement - Prevents marketing language and ensures factual descriptions
- Tool Safety Protocols - Blocks prohibited operations and validates parameters
- User Preference Management - Stores and applies communication and aesthetic preferences
- Intelligent Context Monitoring - Automatic token estimation and threshold warnings
- Multi-MCP Orchestration - Coordinated workflows across multiple servers
Active Servers
Enhanced Architecture Server (enhanced_architecture_server_context.js)
Primary server with complete feature set:
- Professional accuracy verification
- Tool safety enforcement
- User preference storage/retrieval
- Context token tracking
- Pattern storage and learning
- Violation logging and metrics
Chain of Thought Server (cot_server.js)
Reasoning strand management:
- Create and manage reasoning threads
- Branch reasoning paths
- Complete strands with conclusions
- Cross-reference reasoning history
Local AI Server (local-ai-server.js)
Local model integration via Ollama:
- Delegate heavy reasoning tasks
- Token-efficient processing
- Hybrid local+cloud analysis
- Model capability queries
Installation
-
Prerequisites:
npm install -
Configuration: Update your Claude Desktop configuration to include the servers:
{ "mcpServers": { "enhanced-architecture": { "command": "node", "args": ["D:\\arch_mcp\\enhanced_architecture_server_context.js"], "env": {} }, "cot-server": { "command": "node", "args": ["D:\\arch_mcp\\cot_server.js"], "env": {} }, "local-ai-server": { "command": "node", "args": ["D:\\arch_mcp\\local-ai-server.js"], "env": {} } } } -
Local AI Setup (Optional): Install Ollama and pull models:
ollama pull llama3.1:8b
Usage
Professional Accuracy
Automatically prevents:
- Marketing language ("revolutionary", "cutting-edge")
- Competitor references
- Technical specification enhancement
- Promotional tone
Context Monitoring
Tracks conversation tokens across:
- Document attachments
- Artifacts and code
- Tool calls and responses
- System overhead
Provides warnings at 80% and 90% capacity limits.
User Preferences
Stores preferences for:
- Communication style (brief professional)
- Aesthetic approach (minimal)
- Message format requirements
- Tool usage patterns
Multi-MCP Workflows
Coordinates complex tasks:
- Create CoT reasoning strand
- Delegate analysis to local AI
- Store insights in memory
- Update architecture patterns
Key Features
- Version-Free Operation - No version dependencies, capability-based reporting
- Empirical Validation - 60+ validation gates for decision-making
- Token Efficiency - Intelligent context management and compression
- Professional Standards - Enterprise-grade accuracy and compliance
- Cross-Session Learning - Persistent pattern storage and preference evolution
File Structure
D:\arch_mcp\
├── enhanced_architecture_server_context.js # Main server
├── cot_server.js # Reasoning management
├── local-ai-server.js # Local AI integration
├── data/ # Runtime data (gitignored)
├── backup/ # Legacy server versions
└── package.json # Node.js dependencies
Development
Architecture Principles
- Dual-System Enforcement - MCP tools + text document protocols
- Empirical Grounding - Measurable validation over assumptions
- User-Centric Design - Preference-driven behavior adaptation
- Professional Standards - Enterprise accuracy and safety requirements
Adding New Features
- Update server tool definitions
- Implement handler functions
- Add empirical validation gates
- Update user preference options
- Test cross-MCP coordination
Troubleshooting
Server Connection Issues:
- Check Node.js version compatibility
- Verify file paths in configuration
- Review server logs for syntax errors
Context Tracking:
- Monitor token estimation accuracy
- Adjust limits for conversation length
- Use reset tools for fresh sessions
Performance:
- Local AI requires Ollama installation
- Context monitoring adds ~50ms overhead
- Pattern storage optimized for < 2ms response
License
MIT License - see individual files for specific licensing terms.
Contributing
Architecture improvements welcome. Focus areas:
- Enhanced token estimation accuracy
- Additional validation gates
- Cross-domain pattern recognition
- Performance optimization