Threatviz👁️ - Turn any CVE into AI Powered Threat modeling

• Agentic AI based Threat Modeling Technique
• Supports STRIDE
• Supports PASTA

Just Insert any CVE then it will turn that to full end to end threat modeling workflow.

---

Visualization

---

Tech Stack

• Langgraph for multi-agent orchestration
• Faiss for the knowledge base and RAG
• Groq and OpenAI for model endpoint
• Huggingface for embeddings
• Streamlit for Web Interface

---

Below AI Models are supported

Model	Status
qwen/qwen3-32b	✔️
meta-llama/llama-4-maverick-17b-128e-instruct	✔️
openai/gpt-oss-120b	✔️
openai/gpt-oss-20b	✔️
openai/gpt-4o-mini	✔️
moonshotai/kimi-k2-instruct-0905	✔️

---

Deployment

• Local Deployment with light weight GUI.
• Deployed and tested using Amazon Bedrock AgentCore Runtime and AgentCore Memory services
• AWS CloudWatch for observability

---

Usage

• Please check .env.example file to setup API key.
• Once setup done, change it from .env.example to .env

git clone https://github.com/findthead/Threatviz.git
cd Threatviz
uv sync
uv run threatviz.py -id CVE-2025-55182 -html_report 

Host it locally with below command

uv run threatviz.py -dashboard

GUI for Local Deployment

---

Cloudwatch Observability (optional)

---

Please provide the citation.

@software{Threatviz,
  author = {Subhay},
  title = {Threatviz: An autonomous multi-agent Threat Modeling Tool},
  url = {https://github.com/findthehead/Threatviz},
  version = {0.0.3},
  year = {2025}
}