Wallbreaker
An autonomous LLM red-teaming framework that deploys agentic tool-calling loops to probe, attack, and evaluate large-language-model safety boundaries.
51Attack presets
71Native transforms
100+Agent tools
400HarmBench behaviors
7HB categories
# Clone the repo$ git clone https://github.com/JailbrokenAI/wallbreaker.git$ cd wallbreaker$ python -m venv .venv && source .venv/bin/activate(.venv) $ pip install -e ".[dashboard,barcodes,stego]"(.venv) $ wallbreaker
Danger
For authorized security testing only. Always obtain explicit written permission before targeting any model, API, or service.
Where to Go
Installation
Clone, create a venv, and install Wallbreaker.
Configuration
Set up profiles, targets, judges, and agents.
First Session
Walk through the TUI and launch your first engagement.
WebUI V2
The unified browser operator surface — executions, runs, workflows, reports.
Agent Loop
Understand the reason→tool-call→observe cycle.
Arsenal
Browse presets, transforms, and tools.
Guides
Step-by-step walkthroughs for payloads, personas, campaigns, and more.
Methodology
Evidence-based red-teaming principles and technique taxonomy.
Glossary
Definitions of terms used throughout the docs.
Last updated on