m-sec-org/BreachWeave

What it solves

BreachWeave is designed for intelligent penetration testing and Capture The Flag (CTF) scenarios. It addresses the challenge of managing complex attack chains where AI agents often struggle with path deviation, cluttered state management, and premature task termination.

How it works

The system uses a multi-agent architecture consisting of three primary roles:

• Manager: Acts as the control plane, handling global orchestration, scheduling solvers, and organizing collaboration.
• Solver: The active execution unit that performs information gathering, vulnerability verification, and exploit chain progression.
• Observer: A supervisory role that monitors execution trajectories to provide lightweight corrections and prevent the agent from drifting off-track or ending the task too early.

To maintain efficiency, the system separates "Ideas" (directions to pursue) from "Memory" (reusable facts and evidence) and employs context compression to remove noise from tool outputs.

Who it’s for

Security researchers, CTF players, and penetration testers who want to automate the discovery and exploitation of vulnerabilities using a coordinated multi-agent AI system.

Highlights

• Multi-Agent Collaboration: Supports concurrent solvers exploring different attack paths under a single manager.
• Runtime Supervision: An observer agent ensures the process remains efficient and prevents premature termination.
• Layered State Management: Separates strategic directions from factual evidence to keep decision-making clear.
• Context Optimization: Uses rewriting and summarization to keep the LLM context high-signal and low-noise.