What happened
Snyk's October/November 2026 documentation updates confirm Agent Behavior Governance is now generally available (open-preview notice removed) and add MCP Governance risk-based policies, letting teams classify MCP servers via allowlist, prompt-injection-in-a-tool risk dimensions, or unofficial-publisher status and preview impact before blocking. Evo also documents the MCP Risk Score combining server behavior, trust, and implementation signals as a policy condition.
Why it matters
These ship real agent-governance controls into Snyk's existing AI Security Platform: risk-based, conditionable policies over MCP servers and generally-available agent-behavior governance, which matter as coding agents pull in third-party tools and skills the security team never reviewed.
Applicability
Snyk Evo/AI Security Platform customers using coding agents should enable risk-based MCP Governance policies and review the now-GA Agent Behavior Governance controls.