Uploaded October 2026 | Updated November 2026, 6 hours ago
GitHub Copilot SDK を使ったエージェントアプリケーション開発を学ぶ、初心者向けライブ配信です。エージェントループを自分で実装することなく、GitHub Copilot のエージェント機能をアプリケーションやサービスに組み込む方法を紹介します。
モデル API を直接呼び出す場合とは異なり、GitHub Copilot SDK を利用することで、会話管理、タスク計画、ツールオーケストレーション、ストリーミング応答、セッション状態管理など、GitHub Copilot のエージェントランタイムを活用できます。
セッションでは、クライアント、セッション、プロンプト、エージェントランタイムの基本概念を解説します。
また、マルチターン会話の実装、リアルタイム応答のストリーミング、アプリケーションの機能をツールとして公開する方法、MCP サーバーによる外部機能との連携、フックや permission handlers を活用したエージェント動作の制御について学びます。
さらに、GitHub Copilot SDK と GitHub Copilot CLI の使い分けについても説明します。GitHub Copilot CLI はターミナル上での開発作業に適しており、GitHub Copilot SDK は GitHub Copilot のエージェント機能を独自のアプリケーションやサービスへ組み込みたい場合に適しています。
GitHub Copilot SDK の事前知識は必要ありません。
0:00:00 オープニング・自己紹介
0:02:17 GitHub Copilot SDKの概要
0:05:34 GitHub CopilotランタイムのRust移行とパフォーマンス向上
0:08:33 エージェントループの基本概念(設定・組み立て・推論・認可・実行・継続)
0:11:04 エージェントのデプロイ戦略
0:17:01 ワークショップ・デモ開始(コンソールRPGでの活用)
0:21:20 環境準備とアクセシビリティレビューの紹介
0:24:29 Copilotクライアントとセッションの実装
0:29:34 レスポンスのストリーミング実装
0:33:05 ローカルツールの作成と呼び出し
0:41:05 MCPサーバーとPlaywrightによる外部機能連携
0:53:29 開発したエージェント(RPG)の実機デモ確認
1:00:24 エンディング
[eventID:27617]
GitHub Copilot SDK を使ったエージェントアプリケーション開発を学ぶ、初心者向けライブ配信です。エージェントループを自分で実装することなく、GitHub Copilot のエージェント機能をアプリケーションやサービスに組み込む方法を紹介します。
モデル API を直接呼び出す場合とは異なり、GitHub Copilot SDK を利用することで、会話管理、タスク計画、ツールオーケストレーション、ストリーミング応答、セッション状態管理など、GitHub Copilot のエージェントランタイムを活用できます。
セッションでは、クライアント、セッション、プロンプト、エージェントランタイムの基本概念を解説します。
また、マルチターン会話の実装、リアルタイム応答のストリーミング、アプリケーションの機能をツールとして公開する方法、MCP サーバーによる外部機能との連携、フックや permission handlers を活用したエージェント動作の制御について学びます。
さらに、GitHub Copilot SDK と GitHub Copilot CLI の使い分けについても説明します。GitHub Copilot CLI はターミナル上での開発作業に適しており、GitHub Copilot SDK は GitHub Copilot のエージェント機能を独自のアプリケーションやサービスへ組み込みたい場合に適しています。
GitHub Copilot SDK の事前知識は必要ありません。
0:00:00 オープニング・自己紹介
0:02:17 GitHub Copilot SDKの概要
0:05:34 GitHub CopilotランタイムのRust移行とパフォーマンス向上
0:08:33 エージェントループの基本概念(設定・組み立て・推論・認可・実行・継続)
0:11:04 エージェントのデプロイ戦略
0:17:01 ワークショップ・デモ開始(コンソールRPGでの活用)
0:21:20 環境準備とアクセシビリティレビューの紹介
0:24:29 Copilotクライアントとセッションの実装
0:29:34 レスポンスのストリーミング実装
0:33:05 ローカルツールの作成と呼び出し
0:41:05 MCPサーバーとPlaywrightによる外部機能連携
0:53:29 開発したエージェント(RPG)の実機デモ確認
1:00:24 エンディング
[eventID:27617]
![dbt job in Fabric: GA ready
The session will focus on the General Availability of dbt Job in Microsoft Fabric and will cover:
Whats new in dbt Job, including key capabilities and enhancements available as part of the GA release.
A look at the product roadmap and future investments to help customers understand whats coming next.
Customer success stories and lessons learned from organizations running dbt Job in production on Fabric today.
0:00:05 Introduction and Housekeeping
0:03:43 What is dbt and its Benefits
0:08:10 dbt in Microsoft Fabric Capabilities
0:12:40 Data Ingestion Options
0:15:05 End-to-End Demo
0:18:02 Partnership with dbt Labs and dbt Fusion
0:19:25 Additional Features
0:23:00 Architecture Patterns in Fabric
0:26:15 Product Roadmap
0:33:10 Customer Success Stories
0:34:07 Cost and Pricing Breakdown
0:44:45 Getting Started and Hands-on Lab
[eventID:27532] dbt job in Fabric: GA ready](https://i.ytimg.com/vi/I6zny-w3mMQ/mqdefault.jpg)
![Ignite 2026: Whats New, Whats Next?
Get ready for a deep dive into Microsoft Ignite 2026: Whats New, Whats Next?
Explore the latest announcements, innovations, and roadmap across Microsoft AI, Copilot, Microsoft 365, AI agents, Azure, Microsoft Foundry, data, and security.
Discover whats changing, whats coming next, and what these innovations mean for developers, IT professionals, and technology leaders.
Well unpack the key highlights, exciting new capabilities, and practical opportunities to start building with Microsofts latest technologies.
[eventID:27795] Ignite 2026: Whats New, Whats Next?](https://i.ytimg.com/vi/IQniX75vRwY/mqdefault.jpg)
![Finalists Revealed: What a Winning Power BI Dataviz Looks Like
These are the names to remember.
Mariia Drobina. Marius Helle. Aanor Le Gouic. Jeromeian Arbonida. They are all headed to the finals.
Meet the Power BI Dataviz competition finalists and take a closer look at the work that earned them a spot on the global stage.
You’ll see how they approached the challenge, what made their visuals stand out, and how they turned ideas into something that rises above the rest. They’re headed to Barcelona to compete live on stage at FabCon.
They’ll have 20 minutes (that’s 1,200 seconds) to create a world-class visual in front of a cheering audience. When it’s over, only one person can call themselves the Power BI Dataviz World Champion.
Pick your favorite and join their virtual cheer squad before they hit the stage.
🔗 Learn more about the World Championships: https://aka.ms/dd/create
📍 This session is a part of a series - learn more here: https://aka.ms/dataydayslive/viz-y
Chapters:
0:00:07 Introduction and Housekeeping
0:00:54 Host Introductions: Valerie and Lakshmi
0:02:43 Meet the Moderators: Johan and Miguel
0:04:01 Overview of the Power BI Dataviz World Championship
0:07:00 Finalists Revealed: Meet the Contestants
0:15:42 Report Showcase: Maria Drobina (Movies Data)
0:24:14 Report Showcase: Aanor Le Gouic (Lego Data)
0:33:11 Report Showcase: Jerome Ian Arbonida (Sports Data)
0:46:00 Report Showcase: Marius Helle (Sales Data)
0:54:13 Rapid Fire Questions Round
1:00:06 Cheer Squads and Final Wrap-up
[eventID:27528] Finalists Revealed: What a Winning Power BI Dataviz Looks Like](https://i.ytimg.com/vi/J3RWKDI_fT4/mqdefault.jpg)
![Durable Execution for Any Agent on Microsoft Foundry
Microsoft Foundry gives you hosting, identity and models for your agents.
This session adds Agents are long running, stateful and expensive. When one crashes halfway through a multi-step task, you lose its progress, its memory and the tokens youve already spent.
This session shows how durable execution makes Microsoft Agent Framework agents production-ready with Dapr and Diagrid: every step is checkpointed, memory survives restarts, and agents communicate over pub/sub. We then deploy to AKS to show why AKS plus Diagrid is a strong platform for durable agents on Kubernetes. as the durable execution layer underneath. You write agent workflows as plain sequential code, the engine handles recovery through deterministic replay, and workflow state lives in an Azure database you choose, such as Azure Database for PostgreSQL.
It works with any agent framework, from MAF in .NET to LangGraph or CrewAI in Python, all running on Foundry with one durable execution engine.
[eventID:27780] Durable Execution for Any Agent on Microsoft Foundry](https://i.ytimg.com/vi/JAoc5YuRhH8/mqdefault.jpg)
![Advanced Extensibility Patterns with WorkIQ Developer Tools
Go beyond the basics and explore advanced extensibility patterns with WorkIQ Developer Tools, from enterprise integrations and grounding to actions and orchestration approaches for richer Copilot experiences.
See how the patterns introduced throughout the series can be applied to real development scenarios, with WIQD resources and cookbook recipes available as practical references when you’re ready to build.
0:00:00 Intro and Housekeeping
0:00:57 Introduction to WorkIQ Developer Tools
0:02:43 Recap of Previous Episodes
0:03:39 Mental Model: Plug-in Package and Components
0:04:50 Connecting to Remote MCP Servers
0:05:48 Authorization Models Overview
0:07:23 Deep Dive into OAuth 2.0 Flow
0:08:26 Step-by-Step Configuration Requirements
0:09:46 Protocol Flow Explanation
0:11:15 Demo: Manual OAuth Configuration
0:22:34 Demo: Building with WorkIQ CLI
0:37:44 Q&A: Entra vs. OAuth
0:41:34 Q&A: How MCP Protocol Works
0:43:24 Q&A: Adoption and Partner Plugins
0:45:34 Q&A: Application-Only vs. Delegated Security
0:53:13 Q&A: Licensing Requirements
0:55:20 Resources and Wrap-up
Get started: https://microsoft.github.io/wiqd
Practical guide: https://aka.ms/wiqd/cookbooks
[eventID:27615] Advanced Extensibility Patterns with WorkIQ Developer Tools](https://i.ytimg.com/vi/JNyNv8U3Urw/mqdefault.jpg)
![Proteção de dados na era da IA
A IA está transformando a forma como trabalhamos, mas também traz novos desafios para a segurança, privacidade e conformidade dos dados. Nesta sessão, vamos explorar como o Microsoft Purview ajuda as organizações a adotar a IA com confiança, permitindo descobrir, classificar, proteger e governar seus dados. Abordaremos cenários práticos envolvendo dados sensíveis, prevenção contra perda de dados (DLP), proteção de informações e governança de IA, mostrando como equilibrar inovação, segurança e conformidade.
🔗 Aprenda mais acessando os recursos:
https://aka.ms/DSPM/y
https://aka.ms/DSPMIA
0:00 Introdução e boas-vindas
1:25 Apresentação do tema: Proteção de dados na era da IA
2:39 Sobre o palestrante: Experiência e certificações
4:50 Panorama: Proteção de dados no Microsoft 365
8:19 Riscos de segurança em agentes de IA
11:03 Pilares de governança com Microsoft Purview
14:35 Ciclo de vida dos dados na utilização de IA
16:20 A importância da segurança na adoção da IA
19:30 Proteção de agentes e Shadow AI
24:16 Integração com outras ferramentas de IA (LLMs)
28:54 Demonstração prática do portal Microsoft Purview
32:51 DSPM: Visibilidade e métricas de risco
34:07 Criação de planos de remediação e políticas de DLP
39:53 Observabilidade de agentes de IA
42:58 Activity Explorer: Auditoria de interações
46:36 Demo: Bloqueio de prompts com dados sensíveis
50:09 Proteção estendida a IAs de terceiros
52:19 Gestão de agentes no Microsoft 365 Admin Center
56:19 Considerações finais e encerramento
#MicrosoftReactor #LearnConnectBuild
[eventID:27539] Proteção de dados na era da IA](https://i.ytimg.com/vi/K73kHXXM-Hs/mqdefault.jpg)
![Prompt Engineering con Copilot: construye prompts que realmente funcionen
¿Utilizas Copilot pero no siempre obtienes los resultados que esperas?
En esta sesión práctica aprenderemos a dejar de improvisar prompts y construiremos juntos un framework sencillo para crear instrucciones más claras, efectivas y reutilizables con Microsoft Copilot.
Trabajaremos con seis elementos: contexto, objetivo, información, restricciones, resultado esperado y criterios de evaluación. A través de ejemplos y demostraciones en vivo, transformaremos prompts simples en instrucciones estructuradas que los participantes podrán reutilizar y adaptar a situaciones reales.
El objetivo no es memorizar “el prompt perfecto”, sino aprender una metodología que permita diseñar, probar y mejorar nuestros propios prompts.
🔗 Aprende más accediendo a los recursos:
https://aka.ms/CopilotCoworkOverview
https://aka.ms/CopilotPrompts109
📌 Este evento es parte de una serie, aprende más aquí: https://aka.ms/BuildWithCopilot/serie
0:05 - Introducción y código de conducta
1:04 - Bienvenida y presentación de la serie
1:33 - Introducción al concepto de prompt
2:39 - Presentación de la ponente (Amal Hosri)
3:49 - Panorama actual de Microsoft Copilot
5:24 - ¿Qué es un prompt y la ingeniería de prompts?
10:42 - Los 6 elementos clave para un prompt efectivo
14:31 - Inicio de la práctica: uso de Copilot en Word
27:03 - Demo práctica: Copilot en Excel
40:34 - Demo práctica: Copilot en Outlook (correo electrónico)
47:46 - Resumen del framework reutilizable
51:33 - Cierre y próxima sesión de la serie
#MicrosoftReactor #LearnConnectBuild #Copilot
[eventID:27606] Prompt Engineering con Copilot: construye prompts que realmente funcionen](https://i.ytimg.com/vi/KIClQdFAuLE/mqdefault.jpg)
![Lo que construimos, lo que funcionó y lo que aprendimos
Rara vez aprendemos más de la parte de un proyecto que salió perfectamente.
En esta sesión, personas de la comunidad compartirán lo que crearon, lo que les sorprendió, lo que salió mal y lo que harían de otra manera la próxima vez.
[eventID:27760] Lo que construimos, lo que funcionó y lo que aprendimos](https://i.ytimg.com/vi/KWWpexJsWn4/mqdefault.jpg)
![How Copilot, Fabric IQ, Fabric Apps, and Your Data Fit Together
There are a lot of new names in the Fabric AI story. Let’s slow down and connect the pieces.
We’ll look at how Copilot, Fabric IQ, Fabric Apps, and your data work together, what each one is for, and where they could fit into the work you’re already doing.
For more information on Data Days and all it has to offer, visit https://aka.ms/DataDays
📌 This event is part of a series. Learn more here: https://aka.ms/AIatWork/UnderstandIt/y
#MicrosoftReactor #LearnConnectBuild #Fabric #AI #DataDays
[eventID:27715] How Copilot, Fabric IQ, Fabric Apps, and Your Data Fit Together](https://i.ytimg.com/vi/Kb0rqQCMKcY/mqdefault.jpg)
![New Year, New Bot - Bring Your AI Desk Robot to Life
Meet the M5Stack StackChan as a physical AI desk robot. Program its expressions and movements, connect it to a .NET backend and dashboard, and create a first end-to-end interaction.
This hands-on session shows how hardware and .NET services can work together to bring a real desk companion to life.
🔗 Learn more by exploring the resources:
https://aka.ms/newyearnewbot
https://aka.ms/M5StackStackChanfirmwaredevelopersetup
https://aka.ms/M5StackStackChanproductinformation
https://aka.ms/dotNETAspiredocumentation
📌 This event is part of a series. Learn more here: https://aka.ms/NewYearNewBotSeries
#MicrosoftReactor #LearnConnectBuild #dotNet #Aspire #Agent #M5Stack #StackChan, #IoT #EdgeComputing
[eventID:27761] New Year, New Bot - Bring Your AI Desk Robot to Life](https://i.ytimg.com/vi/L-zaWiyjDEU/mqdefault.jpg)
![De MCP a agentes A2A en Microsoft Fabric: Fabric MCP en acción
En esta sesión conectaremos un agente o cliente de IA con Microsoft Fabric mediante MCP para entender cómo los agentes descubren capacidades, consultan metadatos y trabajan con artefactos de Fabric.
Explicaremos la diferencia entre MCP local y remoto, revisaremos cómo se autentican las llamadas, qué permisos intervienen y qué tipo de operaciones puede realizar un agente sobre un workspace.
La sesión tendrá un enfoque práctico para mostrar cómo MCP se convierte en la primera capa de conexión entre agentes y Fabric.
🔗 Aprende más accediendo a los recursos:
https://aka.ms/FabricMCPservers
https://aka.ms/FabricCoreMCPServertools
https://aka.ms/FabricCoreMCPServer
https://aka.ms/SkillsforMSFTFabric
https://aka.ms/InstallSkillsforFabric
https://aka.ms/A2AandMCPComparison
https://aka.ms/GitHubSpecKit/y
https://aka.ms/Codexbestpractices
https://aka.ms/OpenAIAgentSkills
📌 Este evento es parte de una serie, aprende más aquí:
https://aka.ms/MCPaA2AFabric/y
#MicrosoftReactor #LearnConnectBuild #MicrosoftFabric #MCP #Agents #OpenAI #Codex #DeveloperTools
[eventID:27821] De MCP a agentes A2A en Microsoft Fabric: Fabric MCP en acción](https://i.ytimg.com/vi/L4L3LZj5dC4/mqdefault.jpg)