Knox – Govern AI agent tool calls before they execute (github.com)

Qoris has launched Knox, a robust security policy engine designed to manage and govern AI coding agents. Knox is available in multiple formats, including a standalone CLI, Node library, and plugins for Claude Code, Cursor, and OpenAI Codex, all built from a single source and shared rule set. The free, open-source Developer Knox protects local development sessions, while the enterprise-grade Qoris Runtime Knox ensures oversight of AI workers operating across various business functions, featuring features like approval workflows, audit pipelines, and shared memory governance. The significance of Knox lies in its ability to enforce security measures in real-time, intercepting tool calls before they execute and preventing potentially harmful actions. With a comprehensive capability matrix, Knox offers programmatic checks, real-time blocking of dangerous commands, and an audit trail of all tool interactions. This dual-layer approach provides developers the flexibility to incorporate Knox for inspection or as a mandatory safety layer, addressing key vulnerabilities in long-running autonomous workflows where AI decisions are less monitored. As AI systems increasingly integrate into operational contexts, tools like Knox enhance both safety and compliance, thereby fostering greater trust in AI technology.