RAMPART
A pytest-native framework for writing and running safety and security tests against agentic AI applications, covering adversarial attacks, benign failures, and a range of harm categories with evaluation-driven assertions.
RAMPART (Risk Assessment & Measurement Platform for Agentic Red Teaming) provides a structured, developer-friendly way to write and run safety and security tests for AI agents. Tests integrate directly with pytest, so assertions about agent behavior read like ordinary test cases rather than a separate evaluation harness.
Test coverage spans adversarial attacks intended to make an agent misbehave, benign failures that occur without any adversarial intent, and a broad range of harm categories, all judged through evaluation-driven assertions rather than manual review.
It is written in Python and released under the MIT license by Microsoft.