Microsoft has merged AutoGen and Semantic Kernel into a single Microsoft Agent Framework targeting general availability in Q1 2026, according to a roundup by AsherTech. The consolidation ends a period of internal fragmentation where two separate Microsoft-backed projects competed for the same developer audience — enterprise teams building AI agent systems.
The merger itself makes organizational sense. AutoGen, a multi-agent orchestration framework from Microsoft Research, and Semantic Kernel, a more structured enterprise SDK, had significant overlap in scope. Running both simultaneously created confusion about which path Microsoft actually recommended for production deployments.
AG2: The Fork That Tells the Real Story
What makes this more than a routine product announcement is what happened next. Original AutoGen contributors, unhappy with Microsoft absorbing their project into a corporate product, forked the codebase and launched AG2 as a community-driven alternative.
The pattern is well-worn. Oracle acquired Sun Microsystems and absorbed MySQL — the community forked MariaDB. HashiCorp changed Terraform’s license — the community created OpenTofu. IBM acquired Red Hat’s CentOS — Rocky Linux and AlmaLinux emerged within months.
Each case follows the same arc: a corporation consolidates an open-source project into its commercial strategy, and a subset of the original contributors decides they’d rather maintain independence than accept corporate governance. The AG2 fork follows this script precisely.
Why It Matters Now
The timing adds an extra dimension. The AI agent ecosystem is in a period of acute tension over trust in centralized platforms. OpenClaw’s CVE-2026-25253 vulnerability exposed the risks of mass adoption outpacing security review. Cloud providers like AWS are rushing to offer managed deployments that promise better defaults. And now Microsoft is consolidating its own agent tooling under tighter corporate control.
For enterprise developers choosing a framework, the Microsoft Agent Framework offers the comfort of a supported, integrated product — likely with Azure-native features, enterprise SSO, and compliance tooling that independent projects can’t match. AG2 offers something different: community governance, transparent roadmap decisions, and freedom from a single vendor’s strategic priorities.
The Broader Signal
The AG2 fork is a small data point in a larger pattern. As AI agent frameworks move from experimental to production, the governance models around them are being stress-tested. Who controls the roadmap? Who decides which features ship? Who gets to change the license?
These questions matter more for agent frameworks than for traditional software libraries because agents operate with elevated permissions — executing code, accessing APIs, managing workflows. The governance of the framework directly affects the trust model of everything built on top of it.
Microsoft’s Agent Framework will likely win the enterprise adoption race on distribution alone. But AG2’s existence ensures that developers who want an alternative have one — and that Microsoft faces at least some competitive pressure to keep its framework open and interoperable.
Sources: AsherTech Agentic AI Tools Guide, AG2, NVD CVE-2026-25253