What happened
Announced June 23, 2026, with GA set for June 29, Snyk's Evo ADS adds a three-layer enforcement plane inside the agent execution loop: (1) agent supply chain security — discovering and scanning MCP servers, skills, and external tools for prompt injection and malicious patterns before agents touch them; (2) runtime behavior governance — real-time policy enforcement blocking destructive agent actions before they execute; (3) trusted output — scanning AI-generated code for vulnerabilities at the moment of creation. Snyk's own telemetry across ~9,700 developer environments shows 50%+ have live MCP server connections and 1-in-12 carry high/critical findings.
Why it matters
This is the first GA-track product from a major DevSec vendor that governs the full agentic development lifecycle — from what tools agents pull in, to what they do at runtime, to what code they ship. It directly addresses the MCP supply-chain attack surface that produced a wave of CVEs this week, and fills a gap that API gateways and static scanners cannot.
Applicability
DevSecOps and AppSec teams deploying AI coding agents (Claude Code, Copilot, Cursor, Codex, Windsurf) should evaluate before GA on June 29. Particularly urgent for enterprises with MCP servers in developer environments.