Decepticon – Autonomous Red Team Agent that Integrates Claude and Codex
Open-source Decepticon red team agent links Claude Code and Codex to a LangGraph runtime for scoped attack-path testing.
Decepticon is an Apache 2.0 autonomous red team agent that lets Claude Code and OpenAI Codex monitor and steer engagements over the Model Context Protocol. Offensive work runs in a LangGraph server inside a Kali Linux sandbox, with 16 specialist agents covering discovery, exploitation, Active Directory, cloud, smart contracts, and reverse engineering. Developers report 102 of 104 XBOW benchmark challenges passed, though the tests included vulnerability-category hints and are not independently verified on enterprise networks. Written permission is required before testing, and a planned defensive retest loop is not yet implemented.