Microsoft's Agent Framework 1.0 Goes Stable and the Agent Governance Toolkit Gets Its Full Architecture Story

Tools

Microsoft Agent Framework 1.0 Is Now Production-Stable for Both .NET and Python

Agent Framework 1.0 locks stable APIs with long-term support, multi-agent orchestration, A2A/MCP protocol support, and first-party connectors for OpenAI, Anthropic, Bedrock, Gemini, and Ollama. If you are still on AutoGen or Semantic Kernel, the migration guides are live and now is the time to move.

Agent Governance Toolkit: Seven Packages, Sub-Millisecond Policy Enforcement, and Full OWASP Coverage

The toolkit ships as seven independently installable packages covering policy enforcement, cryptographic agent identity, runtime supervision, SRE practices, compliance grading, plugin security, and RL training governance. It maps to all 10 OWASP agentic AI risks and integrates with LangChain, CrewAI, Google ADK, OpenAI Agents SDK, and more without requiring rewrites.

Agent Framework Python 1.0.0 Release Notes: Breaking Changes and What Actually Changed

The 1.0.0 release removes deprecated aliases and the old Message constructor parameter, which will break RC installs if you do not read the migration guide first. The changelog is the fastest way to audit your upgrade path before bumping the version.

News

What the Agent Governance Toolkit Actually Contains and Why Security Teams Should Care

Help Net Security breaks down the governance stack from a security practitioner angle, focusing on the cryptographic identity layer, dynamic trust scoring, and the compliance automation that maps to EU AI Act, HIPAA, and SOC2. Worth reading if you need to explain the security posture of agent deployments to someone who does not live in the AI stack.