ToolExperimental

Agent-Wiz

Name: Agent-Wiz
Rating: 3.0 (383 reviews)
Author: Repello-AI

by Repello-AI

CLI threat modeling and visualization for multi-agent systems

Python

Updated Nov 2, 2025

383

Stars

Forks

View on GitHub

Overview

Provides a CLI for threat modeling and visualizing AI agent architectures to uncover attack surfaces and failure modes. Parses agent definitions from popular agent frameworks and generates diagrams, threat lists, and mitigation notes to help teams reason about agent interactions. Includes exports and integrations for common stacks (langgraph, autogen, crewai, swarm) so you can evaluate real project manifests quickly Dynamic Task Routing Pattern. This workflow is informed by the LLM-as-Judge Pattern to ground evaluations.

Key Benefits

As multi-agent systems grow, systemic failures and emergent attack paths become harder to spot; threat modeling shifts security left by making those interactions explicit. Agent-Wiz helps surface agent-to-agent risks Agent-to-Agent Protocol (A2A) and reproducible findings that feed into continuous agent evaluation and reputation tracking. That visibility is critical for building agent trust signals and for connecting evaluation results to operational controls or reputation systems.

Ideal For

Security engineers and platform teams who need a fast way to map, analyze, and document risk in multi-agent deployments. This is especially valuable for teams adopting the Open Agent Specification (Agent Spec) OPEN AGENT SPEC.

Real-World Examples

Identify attack surfaces in agent orchestration graphs generated by langgraph or autogen
Document agent delegation chains and surface failure modes for security reviews
Export threat findings to incident or governance workflows (e.g., n8n) during pre-production testing

Explore evaluation patterns

Methods to assess this tool