mcp-server-semantic-analysis

Semantic Analysis MCP Server

A powerful multi-agent semantic analysis system built with the Graphite framework, providing comprehensive code and conversation analysis capabilities through multiple interfaces.

Overview

This MCP server implements a sophisticated 7-agent architecture for semantic analysis:

• Coordinator Agent - Workflow orchestration and quality assurance
• Semantic Analysis Agent - Core LLM analysis with multi-provider fallback
• Knowledge Graph Agent - Entity and relationship management with UKB integration
• Web Search Agent - Context-aware search and validation
• Synchronization Agent - Data sync across MCP Memory, Graphology DB, and shared-memory files
• Deduplication Agent - Similarity detection and entity merging
• Documentation Agent - Automated documentation generation

Features

Multi-Interface Access

• MCP Server - Direct integration with Claude Code
• HTTP API - REST endpoints for VSCode CoPilot extension
• CLI - Command-line interface (sal command)

API Key Flexibility

3-tier fallback system for maximum compatibility:

1. ANTHROPIC_API_KEY (Claude) - Primary
2. OPENAI_API_KEY (OpenAI) - Secondary
3. OPENAI_BASE_URL + OPENAI_API_KEY (Custom OpenAI-compatible) - Tertiary
4. UKB-CLI fallback mode (no AI) - Final fallback

Advanced Capabilities

• Workflow Orchestration - Complex multi-step analysis workflows
• Quality Assurance - Agent output validation and auto-correction
• Event Sourcing - Durable workflow state and recovery
• Cross-Directory Execution - Works from any directory
• Incremental Analysis - Delta analysis since last run
• Knowledge Synchronization - Multi-system data consistency

Installation

This system is automatically installed as part of the main coding tools:

# Install the entire coding system (includes this semantic analysis server)
./install.sh

Usage

Command Line Interface

# Interactive semantic analysis
sal

# Repository analysis
sal --repository /path/to/repo

# Conversation analysis  
sal --conversation /path/to/conversation.md

# Incremental analysis since last run
sal --incremental

# Pattern extraction
sal --pattern "architectural-patterns,design-patterns"

# Check workflow status
sal --status

# Get help
sal --help

MCP Tools (Claude Integration)

• determine_insights - Analyze repository or conversation for insights
• analyze_repository - Extract patterns and architectural analysis
• update_knowledge_base - Sync insights to knowledge systems
• lessons_learned - Extract lessons from code or conversations

HTTP API (CoPilot Integration)

RESTful endpoints available at http://localhost:8765 when running:

• POST /analyze/repository - Repository analysis
• POST /analyze/conversation - Conversation analysis
• POST /workflows/start - Start custom workflow
• GET /workflows/{id}/status - Get workflow status

Architecture

Agent Responsibilities

1. Coordinator - Manages workflows, coordinates between agents, performs QA
2. Semantic Analysis - Core LLM-powered analysis with provider fallback
3. Knowledge Graph - Entity/relationship management, UKB integration
4. Web Search - Context gathering and validation
5. Synchronization - Data consistency across storage systems
6. Deduplication - Similarity detection and entity merging
7. Documentation - Auto-generated reports and documentation

Data Flow

User Request → Coordinator → Workflow Engine → Agents → QA Validation → Knowledge Sync → Results

Storage Systems

• MCP Memory - Session-based memory for Claude integration
• Graphology DB - Graph database for CoPilot integration
• Shared Memory Files - Persistent JSON files for team sharing

Configuration

Configuration is handled through environment variables:

# API Keys (3-tier fallback)
ANTHROPIC_API_KEY=your-anthropic-key
OPENAI_API_KEY=your-openai-key  
OPENAI_BASE_URL=your-custom-endpoint  # For custom OpenAI-compatible APIs

# Paths
CODING_TOOLS_PATH=/path/to/coding/repo

# Optional: Custom configuration
SEMANTIC_ANALYSIS_CONFIG=/path/to/config.json

Development

Setting up Development Environment

# Clone the repository (if developing standalone)
git clone <repository-url>
cd mcp-server-semantic-analysis

# Create virtual environment
python -m venv venv
source venv/bin/activate  # or `venv\Scripts\activate` on Windows

# Install dependencies
pip install -e .[dev]

# Run tests
pytest

# Format code
black .
ruff check --fix .

Adding New Agents

1. Create agent file in agents/ directory
2. Implement using Graphite framework patterns
3. Register with coordinator in config/agent_config.py
4. Add tests in tests/agents/

Adding New Workflows

1. Create workflow file in workflows/ directory
2. Define as Graphite Assistant
3. Register in coordinator's workflow engine
4. Add tests and documentation

Integration

This semantic analysis server integrates with:

• Claude Code - Via MCP server protocol
• VSCode CoPilot - Via HTTP API and bridge
• UKB Tools - Direct integration and fallback
• Knowledge Management System - Bi-directional sync
• Git Repositories - Direct analysis capabilities

Troubleshooting

Common Issues

1. API Key Issues: Check the 3-tier fallback chain
2. Port Conflicts: System uses intelligent port management
3. Permission Issues: Ensure proper file permissions
4. Memory Issues: Large repositories may need increased limits

Logging

Comprehensive logging available at multiple levels:

• Agent-specific logs
• Workflow execution logs
• API request/response logs
• Error and debugging logs

Contributing

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Ensure all tests pass
5. Submit a pull request

License

MIT License - See LICENSE file for details.