Assay

Assay is an open-source policy-as-code enforcement and evidence tool designed for MCP (Model Context Protocol) AI agent/security workflows. It acts as a deterministic, fail-closed gate between agents and tools, enforcing what tools can do, generating offline-verifiable audit evidence for every action, and maintaining transparency on decision-making. Assay is built for developers, security engineers, and teams needing robust, reviewable security and audit trails for agent/tool calls in CI and production environments, with kernel-level enforcement on Linux and no hosted backend required.