Opfor

•
588
•

Performs open-source adversary emulation to red-team AI agents, LLM applications, and MCP servers against various security standards.

Opfor is an open-source adversary emulation tool designed to thoroughly test AI agents and MCP servers. It red-teams the full AI agent surface, including prompts, tools, memory, and multi-turn reasoning, by generating targeted attacks for established standards like OWASP LLM Top 10, OWASP Agentic AI Top 10, OWASP MCP Top 10, OWASP API Security, and EU AI Act bias suites. It offers versatile execution methods, from a CLI for engineers to a browser extension for non-developers and an SDK for programmatic integration, ensuring comprehensive and auditable security testing.

Key Features

01Comprehensive OWASP coverage (LLM Top 10, Agentic AI Top 10, MCP Top 10, API Security Top 10)
02Browser extension for no-code red-teaming by non-developers
03Trace-aware testing integrating with observability stacks like Langfuse and Netra
04Autonomous red-teaming with `opfor hunt` for adaptive attack campaigns
05Multiple interfaces: CLI, Browser Extension, MCP Server, IDE Skills, and SDK
067 GitHub stars

Use Cases

01Security testing of AI agents, LLM apps, and MCP servers against known vulnerabilities
02Integrating AI security testing into CI/CD pipelines
03Empowering product managers and QA to perform security testing without writing code