RedAmon is an autonomous AI penetration testing framework that maps attack surfaces into a graph, autonomously exploits them from a Kali sandbox with human approval gates, and opens pull requests to fix discovered vulnerabilities.
The tool solves the problem of manual, time-consuming security assessments by chaining reconnaissance, exploitation, and post-exploitation into a single autonomous pipeline. It builds an attack surface graph, runs exploits against discovered targets within a sandboxed Kali environment, and then triages findings to generate code fixes. Human oversight gates are built into critical decision points, ensuring that autonomous actions remain under human control. The framework integrates with Model Context Protocol (MCP) in both directions: it can consume any MCP server as a tool for its own operations, or be driven as an agent from Claude Code or other external agents.
RedAmon suits organizations running bug bounty programs, red teams, or continuous security assessments who want to reduce manual effort while maintaining control over exploitation and remediation. It is designed for self-hosted deployment, making it appropriate for teams that need to keep security testing within their own infrastructure. The tool's ability to automatically generate and open pull requests with fixes distinguishes it from traditional penetration testing frameworks that stop at vulnerability discovery. It works best in environments where developers can act on automated remediation suggestions and where the attack surface can be meaningfully represented as a graph structure.
The project shows active development with regular commits addressing core functionality and bug fixes. The codebase demonstrates ongoing refinement of the exploitation pipeline and integration points. Community engagement is maintained through multiple channels including Discord and Telegram, indicating sustained attention to user feedback and questions.
