Bulwark

Bulwark is an open-source governance layer for AI agents, designed to sit between autonomous agents (such as Claude Code, Codex, OpenClaw, or any MCP/HTTP client) and external tools/APIs. It enforces YAML-based security policies, manages credentials securely by keeping secrets hidden from agents, inspects data flows for sensitive content or prompt injection, provides full audit logging, and rate limiting. Bulwark is especially useful for teams deploying AI agents that require granular access control, auditability, and secure integration in enterprise or regulated contexts.