MCP Server · ThreatModeler Nexus
Built-in MCP Server

Threat modeling, inside the tools your teams already use.

The built-in MCP Server connects ThreatModeler Nexus to the IDEs, agents, and pipelines where systems are built and changed. The threat model travels with the work, and so does the governance.

Security findings don't reduce risk until they're acted on. We connect both sides.

How it works

Every change, checked against the model.

As work happens, the MCP Server reads each change and checks it against the threat model. When a change opens a new risk, it says so in the workflow, not in a review next week.

Then it hands the developer the fix to make, right in the IDE. Secure design becomes part of the flow of work, not a gate at the end of it.

Because the model updates as code ships, a closed ticket means a threat that's genuinely mitigated, not just marked done, and leaders get a live view of risk across every repository.

Every check runs against the Secure Design Graph
Reads the change. Sees the pull request or commit as it lands.
Checks the model. Compares it against the current threat model for the system.
Flags a review. Tells you when a change needs a fresh security look, and why.
Returns the fix. Hands back the threat, the requirement, and the mitigation to apply.
Stays governed. Runs under the same access and policy as the rest of the platform.
What the agent reads

The context that makes the threat model useful.

The MCP Server doesn't just check code against a generic set of rules. It reads the specific context of your system, the architecture, the IaC, the configuration, and the threat model — so findings are relevant to how the system is actually built, not what the model assumes.

Because it reads from the Secure Design Graph, a control that was never in the design is a fact the agent can detect. The model knows what should be there. The agent finds what isn't.

Source code and APIs. The structure of the application as it exists in the repository: services, endpoints, data flows, and the dependencies between them.
Terraform and IaC. Infrastructure definitions from Terraform, CloudFormation, and Azure Resource Manager, the cloud resources and the configuration your systems actually run on.
Docker and deployment configs. Container definitions, orchestration manifests, and service mesh configurations that shape how components are isolated and communicate at runtime.
Application architecture. Design docs, architecture decision records, and existing threat models, the intended design that the Secure Design Graph records and keeps current.
Existing threat models. Prior threat modeling work that becomes part of the Secure Design Graph, so the agent builds on what's already been decided rather than starting from scratch.
Organizational security policies. Your company's specific risk appetite, approved controls, and design standards: so findings reflect what's right for your system, not a generic checklist.
Four ways teams use it

One server, every stage of the work

However your teams build, the threat model meets them there.

Earlier

From documents

Turn PRDs, architecture docs, and policies into living threat models before a line of code exists.

Seamless

In the IDE

Developers create and maintain models from the repository, inside the AI coding tools they already use.

Consistent

In CI/CD

Every pull request is checked against the model automatically, with governance enforced on every change.

At scale

Across the portfolio

Security leaders query models, requirements, and control gaps to focus investment where it cuts the most risk.

Datasheet

Every use case, in detail

The MCP Server datasheet walks through each use case, the integrations it supports, and the governance behind them.

Read the datasheet
Works with your stack

Plugs into where you already work

Connect the coding tools, source control, and models your teams use today.

AI coding tools
Claude CodeCursorVS Code with CopilotWindsurf
Source control & CI/CD
GitHubGitLabAzure DevOpsJenkins
Bring your own model
ClaudeChatGPTOther frontier models
Not a do-it-yourself connector

Enterprise-grade, not bolted on.

Building your own MCP connection means owning the authentication, the access control, and the audit trail yourself. Most MCP servers were never built for that.

By one 2025 review, just 8.5% of public MCP servers implement OAuth correctly. Ours carries your identity end to end, so the same access that governs a request in the IDE carries through to the rules engine. The AI never steps outside the guardrails you set.

OAuth, done right. Standards-based authorization, not hard-coded keys.
Role-based access. The same permissions that govern your teams govern the AI.
Your single sign-on. Identity follows the SSO and SAML you already run.
A full audit trail. Every request and change recorded and defensible.

See what could go wrong, before it does.