Guide public

Run Threat Modeling Sessions with the MCP Gateway

Use BlackShield's governed MCP gateway to run structured STRIDE threat-modeling sessions, attach generated reports to Security Reviews as tenant-scoped evidence, and configure gateway controls so only approved clients can trigger sensitive modeling tools. Audience: Security architects, AppSec leads, tenant admins, and AI Gateway operators. Temps moyen de mise en place: 20-30 minutes.

quickstart

Utilisez ceci si

Use BlackShield's governed MCP gateway to run structured STRIDE threat-modeling sessions, attach generated reports to Security Reviews as tenant-scoped evidence, and configure gateway controls so only approved clients can trigger sensitive modeling tools.

Public visé: Security architects, AppSec leads, tenant admins, and AI Gateway operatorsDurée estimée: 20-30 minutes

Avant de commencer

  • Confirm you have admin access to AI Gateway policy configuration and can edit the policy pack for the target client.
  • Prepare the system architecture description, component inventory, and business context inputs that will seed the STRIDE session.
  • Identify which Security Review record the threat-model output should be attached to after the session.

Parcours rapide

Copiez une configuration de démarrage, exécutez-la dans votre environnement, puis revenez ici pour les détails avancés.

Demonstration only

This configuration is designed for ease of use. To deploy scanner clients at scale, please plan your deployment architecture accordingly or contact us for enterprise best practices.

Exécuter

Minimal AI Gateway policy to enable threat modeling tools

json
{
  "allowed_tools": [
    "threat_modeling.session.create",
    "threat_modeling.session.export",
    "threat_modeling.session.get"
  ],
  "hidden_tools": [
    "threat_modeling.agent_workspace_review.start"
  ],
  "human_approval_required": [
    "threat_modeling.agent_workspace_review.start"
  ],
  "approved_providers": ["anthropic", "openai"],
  "mode": "advisory"
}

Comprendre et personnaliser

Suivez les étapes ci-dessous pour adapter le déploiement, vérifier les accès et étendre l'infrastructure en toute sécurité.

1

Étape 1

Grant threat_modeling.* access in AI Gateway policy

Threat-modeling MCP tools are gated behind explicit AI Gateway policy. Enable them only for approved clients.

  • In `/ai-gateway`, edit the policy pack for the client or client group that will run threat-modeling sessions.
  • Add `threat_modeling.*` to the `allowed_tools` list, or enable individual tools such as `threat_modeling.session.create` and `threat_modeling.session.export`.
  • Use `hidden_tools` to suppress threat-modeling tools from clients that should not see them.
  • Require human approval for `threat_modeling.agent_workspace_review.start` to prevent unsupervised code-reading runs.

Critères de réussite

Require human approval for `threat_modeling.agent_workspace_review.start` to prevent unsupervised code-reading runs.

Open AI Gateway · Connexion requise
2

Étape 2

Start and run a hosted STRIDE session

Hosted sessions run structured STRIDE analysis against architecture inputs through the governed MCP endpoint.

  • Call `threat_modeling.session.create` with your system architecture description, business context, and component inventory.
  • Use the AWS Labs threat-modeling tools exposed by the gateway to step through Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, and Elevation of Privilege.
  • Capture intermediate outputs with `threat_modeling.session.export` at the end of each category for incremental review.
  • Review AI Gateway activity logs after the session to confirm tool calls, latency, and any policy warnings.

Critères de réussite

Review AI Gateway activity logs after the session to confirm tool calls, latency, and any policy warnings.

Open AI Gateway · Connexion requise
3

Étape 3

Attach the session output to a Security Review

Threat-model outputs are persisted as tenant-scoped artifacts and linked to Security Review records for audit and due-diligence.

  • Exported Markdown and JSON session outputs are stored in a tenant-isolated S3 path keyed to your workspace and session ID, encrypted with KMS.
  • Open `/security-reviews` and attach the exported session artifact to the relevant review.
  • Include the threat-model export in the Security Review evidence package when responding to vendor questionnaires or audit requests.
  • Set a data retention period in company integrations settings (default 30 days) to match your policy requirements.

Critères de réussite

Set a data retention period in company integrations settings (default 30 days) to match your policy requirements.

Open Security Reviews · Connexion requise

Critères de réussite

  • The gateway client can invoke threat_modeling.session.create and completes a full STRIDE walkthrough with exported Markdown and JSON outputs.
  • Session artifacts appear in the tenant-isolated S3 path and are attached to the correct Security Review record for audit and due-diligence evidence.
Run Threat Modeling Sessions with the MCP Gateway | BlackShield Docs