Open-Source Security Gets an Upgrade With PentestCode’s Multi-Agent Hacking Engine

A new generation of automated cyber threat defense is now available. Departing from inflexible scanners based on a set of rules for just cross-referencing databases of known vulnerabilities, an open-source AI agent called PentestCode has become a part of the offensive security landscape. 

Refactored by the programmer Zhangir Ospanov into a hard fork of OpenCode, this terminal framework was designed to conduct an autonomous execution of complex campaigns of ethical hacking. Instead of using only one language model to perform an evaluation of an entire network, PentestCode utilizes a unique structure of strategist/coordinator. 

This approach entails splitting intensive cybersecurity tasks between 13 different subagents responsible for particular roles. Upon deployment on a target domain, the coordinator subagent assigns specialized subagents to perform the parallelization of initial data collection and system mapping. Then, once vulnerabilities have been found, subagents specializing in advanced exploitation techniques, including Active Directory credential spraying and Kerberos roasting, are utilized.

The true strength of the platform lies in the native orchestration of the tooling itself. PentestCode automatically runs and performs the actions of 18 different security tools, picking the most appropriate program depending on the context within the feedback loop of the environment. 

For example, if a first-pass port scan identifies the presence of a domain controller, then the agent switches the task and tries to perform local privilege escalation, storing any extracted credentials in the secure chain of evidence. The “assess, decide, execute” strategy results in a claimed increase in processing efficiency by 4.3 times compared to standard single-agent software.

Given that the open-source terminal engine actively creates its own operational memory and checks for active vulnerabilities in the system, this technology can be referred to as dual-use. Not only can authorized red teams within the enterprises use this tool to quickly identify any network exposures, but it can equally well be used by malicious agents against production networks.