ThreatBook has launched RedAgents, a cloud-based platform that connects multiple AI-powered penetration-testing agents to assess the security posture of an organization. Made public on October 16, 2026, its goal is to help security teams uncover vulnerabilities across their networks through automated tests run in parallel—always within pre-approved authorization. The platform centers on breadth over reliance on a single testing tool or provider.
Multi-Agent Approach and Key Players
RedAgents serves as a centralized system to run several AI agents side by side against defined and approved targets. Its promise is to reveal issues that could be missed when using a solo agent. Everything—from test findings to logs—is consolidated into a single interface, simplifying workflow for defenders who have historically juggled multiple tools.
At launch, four distinct agents are part of the platform. Artex Pro is the enterprise-grade version of the autonomous pentesting agent Artex. Cyber(YanJi) operates in Python and is tasked with finding routes from an initial point to a prescribed goal. Yanxun manages coordination among specialist agents, while CyberStrikeAI Enterprise Edition comes from ThreatBook’s most recent acquisition in September 2026. Some performance claims include Yanxun’s ability to orchestrate dozens of agents and Cyber(YanJi) solving all 54 challenge problems in a validation exercise. However, these figures stem from vendor-reported data rather than independent benchmarking.
Controls, Data Handling, and Limitations
ThreatBook asserts that RedAgents accommodates tight operational guardrails. Organizations can specify an allowlist of targets, enforce rate limits, and define timing constraints. Sensitive test data is kept inside ThreatBook’s environment. A Zero Data Retention (ZDR) option lets users request full erasure of data after use.
Despite its advantages, RedAgents doesn’t yet provide public comparative metrics, such as false-positive rates or side-by-side agent performance across the same target. Nor has the platform disclosed customer-case studies or measured outcomes publicly. These are critical for organizations seeking evidence before integrating AI-based pentesting into mature security operations.
The rise of RedAgents aligns with a trend toward automating more cybersecurity tasks using AI agents. Similar offerings include tools from PentestCode or verification-focused platforms like Xalgorix, each varying in deployment strategy, evidence quality, and data governance. RedAgents distinguishes itself by aggregating multiple AI testers into one managed system, rather than offering one single agent or scanning tool.
RedAgents is available via trial requests on ThreatBook’s site. More agents—including open-source—and enhanced features are expected to roll out over time.
Why it matters: As enterprises grapple with an ever-expanding attack surface, automated pentesting powered by AI agents offers scalability and speed. However, the value depends heavily on trust—accuracy of findings, validity of test logs, clarity on data retention, and how well the system adheres to the defined scope. RedAgents’ multi-agent model ups the game, but organizations must validate vendor claims, pilot carefully, and retain skilled teams to evaluate and act on what’s uncovered. Keep an eye on how real-world benchmark data emerges and how this tool adapts to increasingly complex regulatory and compliance pressures.
