ThreatBook Launches AI-Powered Pentest Aggregator, RedAgents

• Global launch of AI agent penetration cluster allows organizations to thoroughly assess their own attack surface before malicious actors can exploit any exposed entry points.
• RedAgents empowers users with multiple leading enterprise-grade proprietary pentest agents — including Artex Pro, Cyber(YanJi), Yanxun, and CyberStrikeAI (Enterprise Edition).

Singapore, Hong Kong, Seoul, Tokyo, Abu Dhabi, Riyadh, London and Palo Alto, 16 October 2026 — ThreatBook, the agentic security company, today launched RedAgents, an AI-powered penetration testing (pentest) aggregator that further bolsters the firm’s Red Team offerings to security operations (SecOps). Orchestrating multiple offensive activities, RedAgents allows SecOps to stage multi-agent attacks against their own networks, with the goal of unearthing hidden and hard-to-detect vulnerabilities, while strengthening their cyber defenses.

Artificial intelligence (AI) is central to the capabilities of RedAgents. Previously, penetration tests were long and laborious to execute, involving significant amounts of manhours and human resources. Today however, high-quality AI penetration agents are fast emerging and constantly evolving — empowering security teams to continuously and regularly undergo penetration testing. The optimal approach, therefore, is to integrate a cluster of AI-powered penetration agents, which continuously advance and are readily available on demand. This ensures a constant supply of the most sophisticated AI pentest technology and the most objective results.

Feng Xue, Founder and CEO of ThreatBook, said: “The overwhelming majority of security teams encounter the same pitfall. Typically, they use a single penetration testing provider, which usually leads to a siloed view of an organization’s security posture. The reality is that no two agents will unearth the same cyber risks: one agent may identify three vulnerabilities, while another agent may identify four, and there could be zero cross-over.”

“No single agent can discover every issue, and RedAgents solves this dilemma for SecOps. By deploying multiple AI pentest agents against the same target, the combined results lead to far higher vulnerability detection rates, while minimizing false positives.”

A noteworthy benefit of RedAgents is the ability to manage every agent in one place and run a variety of agents in a single pentest. RedAgents is a software-as-a-service (SaaS) platform that brings together highly varied AI red team agents from universities, security vendors and open-source communities, and runs them in parallel. Security teams don't need to switch between multiple systems or integrate with numerous vendors. By logging in just once, they can deploy multiple AI red teams on the same penetration test, breaking free from the silos that typically exist between separate tools.

Combining agents is a choice, and each team decides how far to take it. Teams can assemble the selection best suited to a given test, or deploy a single agent, matching the strengths of each red team to their specific use case.

Test data, vulnerability results, and log reports from all agents are displayed in a unified view, eliminating the need for cross-platform consolidation, and reducing operational and analysis costs. RedAgents guarantees that all test data is retained within ThreatBook’s environment, and is not transferred to any third-party agents. Users can define the scope of testing via a whitelist, and the platform operates strictly within authorised parameters. The platform supports Zero Data Retention (ZDR), enabling the complete deletion of test data.

RedAgents empowers users with multiple leading enterprise-grade proprietary pentest agents, including:

• Artex Pro (Enterprise Edition). An extension of Artex (Open-Source Edition), which is an autonomous pentest agent developed as a personal project, by the product lead of RedAgents, Artex Pro enhances security control capabilities by infusing autonomous agents into all-encompassing red team system. Not only can it autonomously breach and discover vulnerabilities, it can also achieve full controllability, stable operation, and traceable results. On 3 September 2026, Artex (Open-Source Edition) won Baidu’s “Agent+ Offensive and Defensive Challenge” hosted by the Baidu Security Response Center. Artex (Open-Source Edition) was officially converted to closed source on 8 October 2026 after being misused by attackers, leading to several large-scale security incidents in the financial services industry. Given the lack of built-in security control measures in Artex (Open-Source Edition), it can only be used cautiously in test environments.
• Cyber(YanJi), a Python-based AI agent system that solves problems by searching through unknown state spaces, was validated first on autonomous penetration testing. Cyber(YanJi) is designed to search for a viable path from a known starting point (origin) to a designated success condition (goal) across an entirely unknown or near-infinite state space. During its validation phase, Cyber(YanJi) demonstrated its efficacy by coming in the 1^st of the individual participants at the “Tencent Cloud Hackathon AI Penetration Testing Challenge”, solving 54 out of 54 targeted problems.
• Yanxun is an intelligent penetration testing system launched by Changsha Yanzhi Technology and based on scientific research conducted by the Peking University Changsha Institute of Computing and Digital Economy. It has repeatedly won attack and defense competitions, including the “Qiangwang Cup National Cybersecurity Challenge”; and continues to deliver results across multiple real-world scenarios. Yanxun holds a distinct advantage in long-penetration scenarios and can simultaneously coordinate several dozens of specialized agents to operate in parallel — thereby reducing the testing cycle from several weeks to a matter of hours. Furthermore, its closed-loop vulnerability data verification mechanism ensures that every vulnerability can be reproduced, with zero false positives. Yanxun supports a wide range of security assessment scenarios, including information gathering, threat intelligence collection, code auditing, penetration testing and red team assessments.
• CyberStrikeAI (Enterprise Version), CyberStrikeAI is an open-source, AI-native autonomous penetration testing tool with native MCP support and integrations with 100+ security tools. It earned over 7,800 GitHub stars within just a few months of launch. CyberStrikeAI Enterprise edition turns security testing requirements into controlled, traceable automated tasks, efficiently orchestrating AI-powered penetration testing and red team exercises within clearly defined, authorized scopes. CyberStrikeAI was acquired by ThreatBook in September 2026.

Post-launch, RedAgents will continue to leverage multiple open-source pentest agents, to complement the capabilities of their enterprise-grade peers, enabling users to experience the latest AI offensive and defensive capabilities into the foreseeable future.

“At ThreatBook, we aspire to break down capability silos in the AI pentest field, by creating a red team open ecosystem that is truly autonomous, controllable, and flexible. SecOps teams don’t need an agent from a single vendor — but rather, a red team that can be deployed at any time and assembled on demand,” XUE added.

Visit RedAgents.ai to find out more about ThreatBook’s latest AI agent penetration cluster and apply for a trial today.

###