Microsoft — Agentic AI Overview
Overview
Microsoft's agentic AI strategy centres on Azure AI services, the open-source Semantic Kernel SDK, and the AutoGen multi-agent framework. Integration with Microsoft 365, GitHub Copilot, and the broader Azure ecosystem makes Microsoft a strong choice for enterprise agent deployment.
Key Offerings
| Product / Area | One-liner | Wiki Reference |
|---|---|---|
| Microsoft Agent Framework | Microsoft's opinionated framework for building production-grade agents on Azure | AgenticFrameworks/microsoft-framework.md |
| Semantic Kernel | Open-source SDK for integrating LLMs into .NET, Python, and Java applications | AgenticFrameworks/semantic-kernel.md |
| AutoGen | Microsoft Research framework for multi-agent conversation and orchestration | AgenticFrameworks/autogen.md |
| Azure AI Agent Service | Managed Azure platform for deploying, running, and monitoring agents | AgentPlatforms/microsoft-azure.md |
| Azure AI Foundry Agent Service | Developer-grade managed runtime with Entra identity, private networking, multi-framework SDK (LangGraph, Claude, OpenAI), scale-to-zero | AgentPlatforms/enterprise-platforms-2026.md |
| Microsoft Copilot Studio | Low-code agent builder for M365/Teams; 160,000 organizations, 400,000+ custom agents (May 2026) | AgentPlatforms/enterprise-platforms-2026.md |
| Agent Governance Toolkit | MIT-licensed runtime governance: deterministic policy enforcement (YAML/Rego/Cedar), zero-trust identity (Ed25519 + ML-DSA-65), execution rings, MCP security gateway; covers all 10 OWASP Agentic risks | SecurityFrameworks/agent-governance-toolkit.md |
| Azure Skills Repository | Official GitHub repository of ready-made skills targeting Azure AI Agent Service and Semantic Kernel toolchains | Standards/skills.md#provider-skills-repositories |
| SkillOpt | Text-space optimizer that trains reusable natural-language skill documents for frozen LLM agents; 52/52 benchmark wins without weight updates | PromptEngineering/skillopt.md |
| AgentPex | Open-source (MIT) agent evaluation tool that imports execution traces, extracts specs from system prompts and tool schemas, and scores behavior using 8 LLM-based evaluation techniques; integrates with Langfuse and Langtrace | EvaluationFrameworks/llm-frameworks.md |
| Azure HorizonDB | Fully managed, PostgreSQL-compatible database for AI-era workloads: elastic storage to 128 TB, 3,072 vCores, hybrid DiskANN + full-text vector search, in-database model invocation via azure_ai extension; Public Preview (Build 2026) | AgentPlatforms/microsoft-fabric-databases-2026.md |
| Rayfin | Open-source SDK + CLI to define and deploy a complete application backend (database, APIs, identity, policies) to Microsoft Fabric in one command; data lands in OneLake; Replit partnership; Public Preview (Build 2026) | AgentPlatforms/microsoft-fabric-databases-2026.md |
| Fabric IQ | Shared context layer for AI agents (Operations Agents GA, Graph GA, Planning GA June 2026); business ontology accessible via MCP to any vendor's agents; unifies historical, real-time, and planning data | AgentPlatforms/microsoft-fabric-databases-2026.md |
| Microsoft IQ | Unified intelligence layer combining Work IQ, Foundry IQ, Fabric IQ, and Web IQ (live web grounding) behind a single SLA-backed retrieval endpoint; GA at Build 2026 | AgentPlatforms/microsoft-fabric-databases-2026.md |
| Azure Cosmos DB MCP Toolkit | MCP-based connector for Cosmos DB + semantic reranking + Agent Memory Toolkit (Cosmos DB + Durable Functions + Foundry models for persistent agent memory); Preview (Build 2026) | AgentPlatforms/microsoft-fabric-databases-2026.md |
| SQL MCP Server | Securely connects Azure SQL data to AI agents and Copilots via Model Context Protocol; Public Preview (Build 2026) | AgentPlatforms/microsoft-fabric-databases-2026.md |
| Microsoft Agent Framework 1.0 | GA at Build 2026; opinionated framework for building production-grade agents on Azure | AgenticFrameworks/microsoft-framework.md |
| Power BI Semantic Models + Fabric OneLake | Mature tabular BI semantic layer now materializing as Delta tables in OneLake via Direct Lake mode, bridging BI semantics with the lakehouse | AgenticTechStack/semantic-data-layer-radar.md |