Bulwark: Open-Source Governance for AI Agents

Bulwark is an open-source governance layer designed to address the security and control challenges associated with AI agents. It acts as an intermediary between AI agents and external tools, enforcing policies, managing credentials, inspecting content, and maintaining a comprehensive audit trail. This is achieved through a combination of YAML-based policy rules, last-mile credential injection, content scanning, and tamper-evident audit logging.

The system supports multiple deployment modes, including an MCP gateway for Claude Code and OpenClaw, and an HTTP forward proxy for Codex and other HTTP clients. This flexibility allows organizations to integrate Bulwark into existing AI agent workflows without significant disruption. The policy engine supports scope-based precedence and hot-reloading, enabling dynamic adaptation to changing security requirements.

Key features of Bulwark include rate limiting, cost tracking, and threat model analysis. These capabilities provide organizations with the tools they need to manage the risks associated with AI agent deployments and ensure compliance with relevant regulations. The open-source nature of Bulwark fosters community contributions and customization, allowing organizations to tailor the framework to their specific needs.

Transparency is a core principle of Bulwark's design. The audit logging system provides a detailed record of all AI agent actions, enabling forensic analysis and incident response. The policy engine allows organizations to define clear and auditable rules for AI agent behavior. By prioritizing transparency, Bulwark promotes trust and accountability in AI agent deployments.