elder-plinius/T3MP3ST

autonomous red teaming platform; multi-agent offensive-security meta-harness

What it solves

T3MP3ST is an offensive security framework designed to automate the process of finding vulnerabilities (zero-days) in various targets. It transforms a standard AI coding agent into a security researcher by providing it with a specialized arsenal of tools and a structured "kill chain" workflow (reconnaissance, exploitation, and reporting).

How it works

The framework acts as a "war machine" bolted around an existing AI agent (such as Claude Code, Codex, or local models via Ollama). It uses a ReAct loop to drive a multi-agent swarm—consisting of operators like the Recon engine, Exploiter, and Infiltrator—to interact with targets. It includes a built-in "Arsenal" of over 35 default tools (expandable to 108) and an MCP server for tool execution. It also features egress-scope containment to ensure the agent only targets authorized systems.

Who it’s for

It is built for authorized security researchers, penetration testers, and educators who want to to automate vulnerability hunting across web apps, CTFs, source code, and embedded/robotics systems.

Highlights

  • Keyless Integration: Connects to local AI agents already on the machine, removing the need for additional API keys or cloud costs.
  • Reproducible Benchmarks: Includes a verify-claims command that re-derives all performance metrics from committed data.
  • Broad Target Coverage: Supports web applications (XBEN suite), CTFs (Cybench), and a coordinated-disclosure pipeline for robotics/OT/embedded OSS.
  • Local-First: Can be run fully offline using Ollama, LM Studio, or vLLM.
  • Egress Containment: Built-in networking restrictions that prevent tools from attacking targets outside the defined scope.