Ai - 1

2026 (7)

March (2)

ai: Foundry Agent Service is GA: private networking, Voice Live, and enterprise-grade evaluations

Microsoft AI Foundry Blog

2026-03-16

Foundry Agent Service is now generally available with several new production features, including end-to-end private networking, Voice Live integration, expanded MCP authentication, GA evaluations with continuous monitoring, and hosted agent deployments in six additional Azure regions.

ai: What’s new in Microsoft Foundry | February 2026

Microsoft AI Foundry Blog

2026-03-06

February 2026 updates add new/updated foundation models (Claude 4.6, GPT Realtime/Audio 1.5, Grok 4.0 GA), advance the Microsoft Agent Framework to Release Candidate, and move Foundry REST API v1 to GA.

February (5)

ai: Microsoft Agent Framework Reaches Release Candidate

Microsoft AI Foundry Blog

2026-02-20

Microsoft Agent Framework has reached Release Candidate for both .NET and Python. Microsoft states the v1.0 API surface is stable and intended v1.0 features are complete on the path to General Availability.

ai: What’s new in Microsoft Foundry | Dec 2025 & Jan 2026

Microsoft AI Foundry Blog

2026-02-19

Microsoft Foundry’s Dec 2025–Jan 2026 updates add several new foundation models, expand audio and fine-tuning capabilities, and introduce SDK changes including an azure-ai-projects v2 beta consolidation.

ai: DPO Fine-Tuning Using Microsoft Foundry SDK

Microsoft AI Foundry Blog

2026-02-13

Microsoft Foundry published a guide on using the Foundry SDK to fine-tune LLMs with Direct Preference Optimization (DPO) to better align outputs with human preferences.

ai: Beyond the Prompt – Why and How to Fine-tune Your Own Models

Microsoft AI Foundry Blog

2026-02-11

Microsoft Foundry argues that enterprise LLM deployments often need behavior-level alignment (consistency, reliability, policy compliance) that prompt engineering and RAG cannot provide, positioning fine-tuning as the mechanism to change model behavior.

ai: Building an AI Skills Executor in .NET: Bringing Anthropic’s Agent Pattern to the Microsoft Ecosystem

Microsoft AI Foundry Blog

2026-02-06

Guest post outlining how to implement an “AI Skills Executor” in .NET using Azure OpenAI and the Model Context Protocol (MCP) C# SDK to replicate Anthropic-style agent patterns.

2025 (5)

December (4)

ai: What’s new in Microsoft Foundry | October and November 2025

Microsoft AI Foundry Blog

2025-12-19

Azure AI Foundry has been renamed Microsoft Foundry. The October–November 2025 update highlights new announcements across agents, models, tools, and related platform improvements.

ai: Foundry IQ in Microsoft Agent Framework

Microsoft AI Foundry Blog

2025-12-07

Foundry IQ Knowledge Bases can be integrated into the Microsoft Agent Framework in about 20 lines of Python. The Azure AI Search Context Provider enables intelligent, multi-hop retrieval so agents can use consolidated knowledge without fragmented pipelines.

ai: Announcing Foundry MCP Server (preview), speeding up AI development with Microsoft Foundry

Microsoft AI Foundry Blog

2025-12-03

Microsoft announced the Foundry MCP Server (preview), building on the Model Context Protocol (MCP) to help AI agents securely connect with apps, data, and systems. At Ignite, Microsoft Foundry also introduced Foundry Tools as a central hub to discover, connect, and manage both public and private MCP tools securely, simplifying integration and expanding interoperability.

ai: ⭐Upcoming Virtual Event⭐ AI Dev Days, Level-Up Your AI Skills with Microsoft Reactor

Microsoft AI Foundry Blog

2025-12-03

AI Dev Days is a two-day virtual event from Microsoft Reactor that highlights the latest innovations across Microsoft Azure, Foundry, and GitHub, focused on modernizing apps, building with agents, and exploring new AI models to help developers level up and connect with experts.

November (1)

ai: Introducing Memory in Foundry Agent Service

Microsoft AI Foundry Blog

2025-11-25

Microsoft's Foundry Agent Service introduces a memory capability so conversational agents can retain context across interactions rather than being stateless. This aims to prevent agents from repeating questions and removes the need for developers to build custom memory solutions using embeddings and external databases.