Microsoft RAMPART and Clarity Open-Sourced

Introduction

Microsoft has introduced two new open-source tools designed to improve safety in agentic AI development: RAMPART and Clarity. As enterprise AI systems move beyond answering questions and begin taking actions across email, CRM, code, and business systems, the risks increase significantly. These tools aim to make AI safety a continuous engineering practice rather than a one-time review.

What’s new

RAMPART for continuous agent safety testing

RAMPART is an open-source testing framework built to help engineering teams turn AI red-team findings and incident learnings into repeatable tests.

Key capabilities include:

• Writing safety tests as standard pytest scenarios
• Running tests in CI pipelines like traditional integration tests
• Focusing on cross-prompt injection and related adversarial scenarios
• Supporting statistical trials for probabilistic LLM behavior
• Reproducing production incidents and validating mitigations over time
• Using composable evaluators to inspect tool use, side effects, and policy boundaries

Microsoft says RAMPART is built on top of PyRIT, its open automation framework for red teaming generative AI systems.

Clarity for early design validation

Clarity is designed to help teams question assumptions before implementation begins. Instead of accelerating coding, it helps teams confirm they are solving the right problem.

Clarity supports:

• Structured conversations around problem definition
• Solution exploration and failure analysis
• Decision tracking for architecture and product choices
• Output stored as human-readable Markdown in a .clarity-protocol/ repo directory
• Use through a desktop app, web UI, or embedded coding agent

This creates a documented decision trail that can be reviewed in pull requests alongside source code.

Why this matters for IT and security teams

For security leaders, developers, and platform teams building AI agents, these tools address two common gaps:

• Design mistakes early in the lifecycle that later become expensive security issues
• Difficulty reproducing AI incidents and verifying that fixes truly hold up

RAMPART helps operationalize AI safety testing in the development workflow, while Clarity helps reduce risky architecture decisions before deployment. Together, they support a more disciplined approach to shipping enterprise AI systems.

Next steps

• Review whether internal AI or agent projects need repeatable safety testing in CI
• Evaluate RAMPART for prompt injection and regression testing scenarios
• Consider using Clarity during architecture and product planning phases
• Incorporate design reviews and incident learnings into ongoing AI governance practices

Microsoft has made both tools available now as open-source projects, giving engineering and security teams practical options for safer agent development.