Show HN: Conduct, open-source guardrails for LLM and MCP tool calls (github.com)

A new open-source project, Conduct, has been launched to provide runtime governance for AI agents, ensuring that all actions taken by large language models (LLMs) and associated tools are subject to a unified policy control. This system features two main components: Conduct Guard, a policy engine that determines whether to block, warn, audit, or inject actions before execution; and Conduct Router, which acts as an LLM proxy routing requests through Guard. This framework allows teams to implement a single governance policy across various AI sessions, allowing for greater consistency and security. The significance of Conduct lies in its focus on proactive governance rather than reactive detection, emphasizing policy enforcement during action execution. With features such as signed configuration, a hash-chained audit log, and compliance packs for various standards like OWASP and HIPAA, it aims to empower teams to maintain compliance and integrity within AI workflows. The system is designed for easy deployment via Docker or Kubernetes, making it accessible for users who wish to self-host or utilize its hosted service. Overall, Conduct sets a new benchmark for safety and accountability in AI-driven operations, enabling developers to manage multiple AI interactions seamlessly and securely.