
OWASP project for AI-assisted threat modeling, helping teams identify, analyze, and document security threats and mitigations in system designs.
Point it at a codebase. Get a code-grounded threat model back — ready to review, challenge, and act on. The parts the code can't see — the business context and the architecture behind it — are where you and your architects come in.
Upload a ZIP or import a GitHub repo, and AI Threat Modeler produces a STRIDE-flavored threat-model artifact: a data-flow diagram, candidate threats, and a risk registry — with links to the exact source files and line numbers whenever the AI agent can ground them in real code. An optional adversarial pass drops findings with no code evidence.
Built for AppSec teams who want threat modeling grounded in the actual code — traceable, reviewable, and always current.
Be clear about what you're getting, because the distinction matters:
file:line.file:line, and claims without code evidence are dropped. The result is concrete, reviewable input your team can act on — the code is the source of truth, and your team brings the judgment to act on it.The value you get out is proportional to what you put in (garbage in, garbage out) and to how seriously your team acts on the findings — verifying the diagram, triangulating claims against the code, and adding the business/design context that lives outside the codebase. See How to use the output for the full workflow, and how to use it as a did-we-do-a-good-enough-job check against an existing hand-built threat model.
You'll need Docker and an API key for one of the supported providers — Anthropic (Claude), OpenAI, or DeepInfra to run open-weight models (Kimi, GLM, DeepSeek).
Set up your environment. Copy the env file and add a random JWT_SECRET:
cp .env.example .env
echo "JWT_SECRET=$(openssl rand -hex 32)" >> .env
Build and start:
docker-compose up -d --build
Open http://localhost:3000, log in with admin / admin, add your API key in Settings, and submit a repo.
👉 New here? Follow the Getting Started guide for the full walkthrough with screenshots-worth of detail.
After a job completes, open the report — a set of artifacts for your review, not a final verdict:
| Tab | Contents |
|---|---|
| Data Flow Diagram | Interactive diagram with trust boundaries; click a node to see related candidate threats and source files. Verify this against your real architecture first — findings built on a wrong diagram are suspect. |
| Threat Model | Candidate STRIDE threats with severity, mitigation, and Location (file:line, snippet, GitHub link when imported from GitHub). Treat each as a hypothesis to verify in code, not a conclusion. |
| Risk Registry | Risks cross-linked to threats, with remediation plans — raw material for your team's decisions and tracked remediation. |
Export to PDF, CSV (Excel-friendly), or JSON, so you can pull the findings into your issue tracker and threat modeling practice of record.
Every job runs a short stage → run flow so the AI understands your deployment, not just your code:
Bring your own key for whichever provider fits your budget and compliance needs — pick the provider and model in Settings:
You can also tune reasoning_effort and toggle the adversarial second pass per instance. See the Settings Reference.
| Guide | For |
|---|---|
| Getting Started | First-time users — from install to first report |
| Importing from GitHub | Public and private repos, branches, limits |
| Settings Reference | Tuning cost, depth, and providers |
| Troubleshooting | When something goes wrong |
| Features | Full capability list |
| Deployment | Running with Docker Compose in production |
| SETUP.md | Local development without Docker |
| CHANGELOG.md | Release history |
Full docs index: docs/. Interactive API spec: http://localhost:3001/api-docs (when the backend is running).
npm test # backend + frontend unit tests
cd frontend && npm run e2e:install && npm run e2e # Playwright end-to-end tests
See SETUP.md for the full dev environment (Node 18+, without Docker).