AGT v3.1 Post-Launch: 14 PRs, Community Rules, and a Production Policy Library

Tools

AGT v3.1 Post-Launch Recap: 14 More PRs, 27 Security Fixes, and 5 Production Policy Sets

In the 48 hours after v3.1 shipped, the team merged 14 additional PRs covering ABAC data-layer policies, a zero-config 'AGT Lite' mode, and a reversibility checker that assesses actions before execution. The 5 tiered production policy sets (startup to defense/ITAR) are the most immediately useful thing here for teams trying to ship governed agents fast.

AGT v3.1.0 Release Notes: Unified CLI, Governance Dashboard, and Quantum-Safe Crypto

The unified 'agt' CLI replaces 12 separate binaries and the new shadow AI discovery scanner finds unregistered agents running in your infrastructure, which is a genuinely useful audit capability. ML-DSA-65 post-quantum signing ships alongside Ed25519 so you're not forced to migrate all at once.

Microsoft Agent Framework 1.0 Is Production-Ready with Stable APIs and Long-Term Support Commitment

The stable 1.0 release locks the API surface and commits to backward compatibility, which matters more than any individual feature for teams deciding whether to build on this stack. The migration assistants for AutoGen and Semantic Kernel teams lower the switching cost significantly.

Analysis

Inside the Agent Governance Toolkit: Seven Packages, Sub-Millisecond Policy Enforcement, All 10 OWASP Agentic Risks Covered

The OS-kernel analogy here is the right mental model: policy enforcement intercepts every agent action before execution at under 0.1ms p99, so governance is in the execution path rather than bolted on after. The intent to move the project to a foundation rather than keep it under Microsoft control is worth watching for teams evaluating long-term dependency risk.