Show HN: Enforra – open-source action governance for AI agent tool calls (github.com)

Enforra has launched an open-source SDK aimed at providing robust action governance for AI agent tool calls, addressing a critical gap in security protocols. While system prompts guide AI behavior, the need for a security boundary becomes apparent when an AI can initiate actions like refunds or data exports. Enforra Core allows developers to establish pre-execution policies for tool calls, ensuring that actions are decided based on defined criteria before they are executed. This local runtime operates without network calls and does not execute tools remotely, giving teams control over agent actions and mitigating risks associated with unexpected behaviors. Significantly, Enforra evaluates a tool call against custom-defined YAML policies and returns one of four decisions—allow, block, require approval, or log only. This enhances clarity in decision-making and provides a framework for auditing actions. Developers can simulate policy scenarios locally and test their applications to catch potential issues before they occur, improving overall compliance and accountability. With an architecture that supports varied use cases and local governance, Enforra facilitates safer interactions between AI agents and the tools they utilize, ultimately contributing to a more secure operating environment in AI/ML applications.