Agent-Wiz
VerifiedPython CLI that extracts agent workflows and runs security assessments.
What is Agent-Wiz?
Agent-Wiz is an open-source command-line utility that analyzes Python projects built with popular agent frameworks. It uses static code parsing to identify agents, tools, and communication patterns, then renders them as interactive graphs.
The same extracted structure feeds into automated threat assessments based on established frameworks such as MAESTRO. Reports highlight entry points, assets, and potential risks without requiring runtime execution.
It is intended for developers who need to document agent flows, researchers studying orchestration patterns, and security teams responsible for reviewing LLM-based applications.
Capabilities
What you can build with Agent-Wiz
Map multi-agent systems
Generate clear diagrams showing how agents call each other and invoke tools across supported frameworks.
Produce security reports
Run automated evaluations that list assets, entry points, and risks according to MAESTRO methodology.
Export structured data
Obtain JSON files containing nodes, edges, and threat findings for further processing or integration.
Install Agent-Wiz
pip install repello-agent-wizpip install repello-agent-wiz- 1Install the package with pip.
- 2Set your OpenAI API key as an environment variable or in a .env file.
- 3Run the extract command with your chosen framework and source directory.
- 4Review the generated JSON graph and threat report files.
- 5Open the visualization output to inspect agent and tool relationships.
Works with
Agent-Wiz: pros & cons
Pros
- +Supports a wide range of current agent frameworks through dedicated parsers.
- +Produces both visual graphs and structured threat reports from the same run.
- +Operates via static analysis so no live execution is required.
- +Offers simple CLI commands plus JSON export for automation.
Cons
- –Currently limited to the MAESTRO threat model; others are still in development.
- –Requires an OpenAI API key even for basic extraction and reporting.
- –Accuracy depends on the quality of the AST parsers for each framework.
Frequently asked questions
It currently covers Autogen, CrewAI, LangGraph, LlamaIndex, OpenAI Agents, and several others listed in the project documentation.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other automation options worth comparing.
