ThreatModel-MCP

Model Context Protocol server for AI-powered threat modeling.

Demo

Related MCP server: Mipiti MCP Server

Setup

1. Install dependencies pip install -r requirements.txt

2. Install Graphviz (optional, for PNG diagrams)

   • Windows: Download from https://graphviz.org/download/

   • Mac: brew install graphviz

   • Linux: sudo apt-get install graphviz

3. Configure MCP client (Claude Desktop, etc.)

   {
     "mcpServers": {
       "threatmodel": {
         "command": "python",
         "args": ["/full/path/to/threatmodel_server.py"]
       }
     }
   }

Available Tools

create_threat_model

Creates comprehensive threat models with components, boundaries, and data flows.

Parameters:

• system_name - Name of the system

• components - Array of system components with types, boundaries, security controls

• boundaries - Trust boundaries with security levels (0-10)

• dataflows - Data flows between components with protocols and classifications

• output_format - "diagram", "pytm_code", "threats", or "full_analysis"

• auto_save - Auto-save files (default: true)

• save_path - Directory to save files (default: current directory)

analyze_security_threats

Performs deep security analysis using multiple frameworks.

Parameters:

• analysis_depth - "basic", "standard", "comprehensive", or "paranoid"

• threat_frameworks - ["STRIDE", "MITRE_ATTACK", "OWASP", "NIST", "CIS"]

• focus_areas - Authentication, data protection, network security, etc.

• compliance_frameworks - ["SOC2", "ISO27001", "HIPAA", "PCI-DSS", "GDPR"]

generate_security_controls

Generates security control recommendations based on threats.

Parameters:

• threats - Array of identified threats

• risk_appetite - "low", "medium", or "high"

• technology_stack - Current technologies (AWS, k8s, etc.)

• prioritization_method - "risk_based", "quick_wins", "compliance_driven"

validate_architecture

Validates architecture against security best practices.

Parameters:

• components - System components to validate

• validation_rules - ["zero_trust", "encryption_in_transit", "api_gateway_pattern"]

• architecture_patterns - ["microservices", "serverless", "hybrid_cloud"]

Component Types

Actors: user, admin, service_account Services: server, api_gateway, microservice, lambda, container Data: database, cache, message_queue, file_storage Infrastructure: load_balancer, firewall, external_service

Protocols & Classifications

Protocols: HTTPS, gRPC, WebSocket, SQL, Redis, S3 API Data Classifications: PUBLIC → INTERNAL → CONFIDENTIAL → RESTRICTED → TOP_SECRET

Auto-Save Features

Generated files (with timestamps):

• SystemName_threatmodel_YYYYMMDD_HHMMSS.png - Diagram (when output_format="diagram")

• SystemName_threatmodel_YYYYMMDD_HHMMSS.dot - DOT source (always)

• SystemName_threatmodel_YYYYMMDD_HHMMSS.py - PyTM code (always)

• SystemName_threatmodel_analysis_YYYYMMDD_HHMMSS.md - Analysis report (when output_format="full_analysis")

Example Usage

Example 1: Codebase Analysis

Prompt: "Create a high level threat diagram of current codebase"

Example threat model diagram generated from a cloned OpenAI Codex codebase, showing multi-layer security boundaries, component classifications, and encrypted data flows between services.

Example 2: Web Application Architecture

Prompt: "A web application where the user interacts with a web server, which in turn communicates with a database server. The web server and database server are outside the user's trust boundary. The user connects to the web application via a browser. The web server handles requests and responses, and the database server stores application data. The trust boundary is around the user only; both the web server and database server are outside this boundary"

Example threat model showing user trust boundary with web and database servers in untrusted zone.

Refer to threat analysis report in assets/Web_Application_System_Threat_Analysis_Report.md

Troubleshooting

Graphviz issues: Verify with dot -V DOT syntax errors: Component names automatically sanitized No Python: Ensure Python in PATH