Show HN: Bulwark – Open-source governance layer for AI agents (Rust, MCP-native) (github.com)

Bulwark, an open-source governance layer for AI agents, has been introduced to address the unregulated access and operations of AI systems like Claude Code, Codex, and others. It sits between AI agents and external tools, enforcing policies, managing credentials, and maintaining an audit trail with a unified governance mechanism. Bulwark employs YAML-based rules for policy enforcement, allowing for precise control over tool access, full credential management without exposure to agents, and robust content inspections for sensitive information. Each action taken by the agents is logged in a tamper-evident SQLite database, enhancing accountability. The significance of Bulwark lies in its foundational role in fostering responsible AI development by curbing the risks associated with unchecked AI agent capabilities. Its features, such as hot-reload policy settings, rate limiting, and an MCP-native architecture, facilitate seamless integration with existing AI infrastructures. This solution not only enhances operational security but also ensures compliance with privacy regulations, ultimately empowering developers to safely harness the power of AI technologies.