Architecture Overview¶
High-level architecture and design of SOLVE-IT MCP Server.
Work in Progress
Comprehensive architecture documentation is coming soon. For now, see Implementation Details.
System Components¶
graph LR
A[MCP Client] -->|STDIO/HTTP| B[Transport Layer]
B --> C[MCP Server]
C --> D[Tool Router]
D --> E[20+ Tools]
E --> F[Knowledge Base Manager]
F --> G[SOLVE-IT Data] Core Components¶
Transport Layer¶
- STDIO Transport: For desktop MCP clients (Claude Desktop, etc.)
- HTTP/SSE Transport: For web clients and Kubernetes deployments
Server Core¶
- MCP Protocol Handler: Implements Model Context Protocol specification
- Tool Registry: Manages 20+ forensic investigation tools
- Request Router: Routes tool calls to appropriate handlers
Data Layer¶
- Knowledge Base Manager: Singleton pattern for optimal performance
- SOLVE-IT Library Integration: Access to techniques, weaknesses, mitigations
- Caching: Sub-second response times
Observability¶
- OpenTelemetry: Metrics, traces, and logs
- Structured Logging: JSON format with correlation IDs
- Health Checks: Kubernetes-ready liveness/readiness probes
Security¶
- Multi-layer Security: Input validation, rate limiting, CORS
- Security Middleware: Request validation and sanitization
- Audit Logging: Complete request/response tracking
Deployment Modes¶
Desktop (STDIO)¶
- Single-user mode
- Direct process communication
- Ideal for: Claude Desktop, Cline, other MCP clients
Server (HTTP)¶
- Multi-user mode
- REST API + Server-Sent Events
- Ideal for: Web applications, Kubernetes, production deployments
Performance Characteristics¶
- Startup Time: ~1 second (with shared knowledge base)
- Query Response: Sub-second for most operations
- Memory Footprint: ~100-200MB
- Container Size: 181MB (Alpine-based)
Technology Stack¶
- Language: Python 3.11/3.12
- Framework: MCP SDK (official Anthropic)
- Web Server: Starlette + Uvicorn (ASGI)
- Validation: Pydantic 2.0
- Observability: OpenTelemetry
- Container: Alpine Linux 3.23
Related Documentation¶
- Implementation Details - Complete technical implementation
- Security Model - Security architecture
- Docker Deployment
- Kubernetes Deployment