MCPFast / Tools / Hexgraph: AI Vulnerability Research for Binaries & Firmware
A self-hosted AI agent decompiles, fuzzes, and verifies exploits in a sandbox, recording findings to a typed graph.
View on GitHub→Hexgraph is a self-hosted AI agent designed for automated vulnerability research on binary executables and firmware. It leverages a combination of decompilation, fuzzing, and exploit verification techniques within a secure sandbox environment. The agent systematically analyzes target code, identifies potential weaknesses, and records its findings in a structured, typed graph for efficient analysis and further development.
Hexgraph automates critical stages of binary and firmware security analysis. It begins by decompiling the input binary or firmware image, making the low-level code more accessible for analysis. Following decompilation, it employs advanced fuzzing techniques to probe for vulnerabilities by feeding unexpected or malformed inputs to the program. Once a potential vulnerability is identified, Hexgraph attempts to verify its exploitability within the isolated sandbox. All discovered vulnerabilities, their characteristics, and associated exploits are meticulously documented and stored in a typed graph database.
Hexgraph is an essential tool for AI builders, security researchers, and developers focused on binary and firmware security. It is particularly valuable for individuals and teams engaged in reverse engineering, penetration testing, and the development of secure software. Developers working with embedded systems, IoT devices, or any application involving compiled code will find Hexgraph instrumental in identifying and mitigating potential security risks before they can be exploited.