Microsoft releases two open-source tools intended to improve security practices in agentic AI development. The company introduces RAMPART, a continuous testing and risk-assessment framework for AI agents, and Clarity, a structured design review tool. Microsoft says the tools come from its AI Red Team, an internal group that stress-tests Microsoft’s own AI systems, and that both tools were used internally before being published. RAMPART is built on top of PyRIT, Microsoft’s existing open-source red-teaming library, and is designed to integrate into developer workflows by fitting into a Pytest-native testing approach. The framework provides mechanisms for writing and running safety and security tests for AI agents. Clarity focuses on bringing more discipline to agent design by supporting structured review processes that aim to identify and address potential safety and security issues earlier in development. Across the reports, the common theme is that Microsoft is offering tooling to help developers evaluate agent behavior and reduce risk during building and iteration, rather than treating security as something addressed only after deployment.