Microsoft has officially transitioned Semantic Kernel, its popular open-source AI framework, into maintenance mode, with all new feature development moving to Agent Framework 1.0, which launched in April 2026. The shift represents a strategic consolidation of Microsoft's agentic AI (artificial intelligence) approach, combining Semantic Kernel's enterprise foundations with orchestration innovations from AutoGen into a unified platform.

What Happened to Semantic Kernel?

Semantic Kernel, which debuted in March 2023, was designed to help developers embed large language models (LLMs) into applications by combining AI models with conventional code through plugins, connectors, and orchestration primitives. The framework shipped for.NET, Python, and Java, and ran on any cloud or on-premises setup, including official Amazon Bedrock connectors, with no requirement to use Microsoft Azure.

The framework gained traction among enterprise teams seeking a structured, type-safe way to build AI-powered applications. However, Microsoft's engineering teams recognized an opportunity to unify two separate approaches into a single, more powerful SDK. On April 3, 2026, Microsoft released Agent Framework 1.0 as the production-ready successor, carrying all new feature investment while Semantic Kernel entered a support-only phase.

Critically, this transition does not force existing users to rewrite their applications. Teams already running Semantic Kernel in production can continue using it through at least April 2027, when Microsoft will provide critical bug fixes and security patches. The company designed Agent Framework specifically so that existing plugins, connectors, and enterprise integrations carry forward without requiring a complete overhaul.

How Does Agent Framework Differ From Semantic Kernel?

Agent Framework consolidates capabilities from both Semantic Kernel and AutoGen, Microsoft's multi-agent orchestration library, into a single development experience. The new framework introduces several architectural improvements while preserving the enterprise-grade features that made Semantic Kernel appealing to regulated industries.

The core components of Agent Framework include autonomous agents that use LLMs to process inputs and generate responses, support for multiple model providers including Microsoft Foundry, Anthropic, Azure OpenAI, OpenAI, and Ollama, and graph-based workflows that connect agents and functions for multi-step tasks. A specialized agent type called Harness can reason through long, multi-step tasks, addressing a gap in Semantic Kernel's capabilities.

Agent Framework also introduces native support for the Model Context Protocol (MCP), an emerging standard for connecting AI agents to external tools and data sources, and agent-to-agent (A2A) interoperability, enabling multiple agents to collaborate within a single system.

What Core Features Carry Forward?

• Functions and Plugins: Named containers of functions that models can invoke, created from native code, OpenAPI specifications, text search implementations for retrieval-augmented generation (RAG), or prompt templates.
• AI Service Connectors: An abstraction layer spanning Azure OpenAI, OpenAI, and Amazon Bedrock, covering chat, text generation, embeddings, image, and audio services across multiple providers.
• Vector Store Connectors: A common interface over vector stores like Azure AI Search and Elasticsearch, with vector search exposable as a plugin for retrieval and semantic search.
• Prompt Templates: Templates combining AI instructions, user input, and function output, usable as the start of a chat flow or as callable plugin functions.
• Filters and Observability: Interception points before and after function invocation and prompt rendering for security controls, responsible AI checks, and telemetry through OpenTelemetry.
• Multi-Agent Orchestration: Support for collaborating specialist agents and a Process Framework for structured business workflows.

Why Does This Matter for Developers?

The transition signals Microsoft's confidence in Agent Framework as the long-term platform for enterprise AI agent development. For developers currently using Semantic Kernel, the message is clear: there is no immediate urgency to migrate, but new projects should target Agent Framework to benefit from ongoing innovation and support.

The consolidation also reflects a broader industry trend toward unified agentic frameworks. Rather than maintaining separate libraries for different use cases, Microsoft is betting that a single, well-designed SDK can serve both simple copilot scenarios and complex multi-agent systems. This approach reduces fragmentation and makes it easier for teams to share knowledge and components across projects.

How to Plan Your Migration Strategy

• Assess Current Deployments: Inventory all applications and services currently using Semantic Kernel to understand the scope of any potential migration and identify which systems are business-critical.
• Evaluate Agent Framework Readiness: Review Agent Framework documentation and test its capabilities against your current use cases to determine if it meets your requirements before committing to a migration timeline.
• Plan Incremental Adoption: Rather than migrating all projects at once, consider building new features and applications on Agent Framework while maintaining existing Semantic Kernel deployments until the support window closes in April 2027.
• Leverage Plugin Compatibility: Take advantage of Microsoft's design decision to make plugins and connectors compatible between frameworks, allowing you to reuse existing code components during the transition.

The transition from Semantic Kernel to Agent Framework represents a natural evolution in Microsoft's AI platform strategy. By consolidating two frameworks into one, the company is simplifying the developer experience while maintaining backward compatibility for existing users. Teams with Semantic Kernel deployments have until April 2027 to plan their next steps, giving them ample time to evaluate Agent Framework and plan migrations at a comfortable pace.