Microsoft Build is back in San Francisco - and this year's session catalog reads like a roadmap for the next 12 months of enterprise AI. Microsoft has published the full session catalog for its Build 2026 developer conference, which will be held in San Francisco on June 2-3. This year, the main themes include developer tools and frameworks, cloud platform and data, model training, agents and apps, responsible AI, and… Windows.
If you're an AI developer, platform engineer, solution architect, or technical leader responsible for taking hosted agents from prototype to production, the agenda was practically built for your team. Below: the sessions worth blocking off, three hands-on workshops to prioritize at Build (including the one Kanwal Khipple is proctoring), and the hardware reveal that's already going viral.
We are excited to see Satya Nadella set the narrative for the entire event - and the next year of Microsoft's developer platform strategy. Microsoft Build 2026 will kick off with a keynote featuring CEO Satya Nadella and other Microsoft leaders discussing how the company is creating new opportunity for developers across our platforms in this era of AI.
Local AI is having a moment, and this is the session that connects the dots between Foundry, Windows, and potentially the new Nvidia-powered hardware debuting at the event. This is one of the sessions Microsoft has highlighted as reflective of what to expect during the developer event. If your team is wrestling with the latency, cost, or privacy implications of running everything in the cloud, this is your concrete path to move inference closer to the user - without abandoning your Foundry investments. Pair this with LAB540 (below) for the full picture: build locally, deploy as a hosted agent, instrument it for production.
What to build next when AI can build everything is the question top of mind for many builders. This is the strategy session for architects and engineering leaders. It's one of the sessions Microsoft is spotlighting as representative of the Build 2026 themes, and the question in the title is the question every engineering org is asking right now. With GitHub Copilot writing more of the code, agents handling more of the workflow, and Foundry doing more of the orchestration - what's the human job?
This session shows how to move beyond generic AI and build agents that actually understand your organization’s work. You’ll see how Microsoft Foundry agents can connect to enterprise context using Work IQ, enabling them to operate on real documents, data, and workflows rather than isolated prompts. You’ll learn how to ground agents in your business context, run them under a secure identity, and evaluate their performance using real-world comparisons. The session also explores how organizations can apply governance controls to ensure agents behave predictably and consistently.
This session walks through the full lifecycle of an AI agent - from early experimentation through to secure, enterprise-grade deployment. It introduces Microsoft’s Foundry Agent Service and Agent Framework, showing how organizations can design, deploy, and manage agents at scale. You’ll see how to move from local prototypes into production environments using identity, networking, and lifecycle management capabilities. The session also highlights how tools like GitHub Copilot integrate into this process to accelerate development and deployment.
This session focuses on how to manage risk and maintain control as AI agents become part of real business processes. It demonstrates how to apply governance and oversight across different frameworks and environments, including both Microsoft and open-source solutions. You’ll learn how to define evaluation criteria, test agents against real-world risks, and apply controls that ensure safety and compliance. The session also explores techniques such as adversarial testing and human-in-the-loop validation to improve reliability.
Agents-as-developers is no longer theoretical. Microsoft has called out this session as one to expect during the developer event, and it's the most concrete look you'll get at how agents will reshape day-to-day Windows app development.
Most engineering teams are great at shipping prototypes and terrible at shipping products. Microsoft has flagged this session as part of the Build 2026 lineup worth watching, and for product-minded engineers it's the missing piece: how do you turn a working agent into a distributable, monetizable offering? Bring your PM. Bring your finance partner. Take notes.
Every agent is only as good as the data behind it. This session, also highlighted by Microsoft as representative of Build 2026's themes, is the data + AI bridge - essential if you're building enterprise-ready agents that need to ground answers in trusted, governed data. If you've ever shipped an agent that confidently hallucinated a number, you know why this matters.
Peter Steinberger, the creator of the AI personal assistant OpenClaw, who recently joined OpenAI, is listed as a featured speaker. So is Scott Hanselman, VP and member of Technical Staff at Microsoft and GitHub. Steinberger brings an OpenAI-meets-Microsoft perspective that's rare on a single stage. Hanselman, as always, is a master class in making complex developer topics feel approachable.
Modern agents fail in ways traditional monitoring can't catch. In this hands-on lab, learn how Microsoft Foundry Observability helps you move from prototype → production with context-specific evaluation suites wired into developer workflows via skills/MCP tooling for hosted agents. Then scale quality with continuous evaluation, trace-linked analysis, and adaptive red teaming—and walk away with a sandbox to explore additional features on your own.
This session focuses on one of the biggest gaps organizations face today: how to control, secure, and manage AI agents at scale. It introduces a “universal control layer” approach that brings governance directly into the development process, rather than treating it as an afterthought. You’ll see how modern tools combine security detection, risk assessment, and policy enforcement into a single framework that applies across agents, models, and tools. The session highlights how capabilities like GitHub Advanced Security can detect AI-specific vulnerabilities early, while control-layer tooling helps enforce consistent policies and safeguards across the entire agent ecosystem.
This hands-on lab shows how to build and run AI agents entirely on local devices, without relying on the cloud. Using Lemonade Server and OpenCode, attendees learn how to leverage modern hardware to run advanced AI models directly on their machines - eliminating the need for APIs, subscriptions, or external data access.
You’ll explore how to design specialized AI agents and orchestrate them using OpenCode, where multiple agents can collaborate to complete tasks. The session demonstrates how local AI infrastructure can be used to power real development workflows, including building and running applications using AI-driven coding agents.
This is the announcement already lighting up X, LinkedIn, and every developer Slack channel — and it's happening at Build.
Nvidia is expected next week to debut the first Windows computers that use its chips as the main processor, sources confirm to Axios. Microsoft's first AI PC push stumbled, but Nvidia's arrival gives it a second chance, this time with the world's hottest chipmaker attached. The chips will appear in Microsoft Surface computers and PCs from other manufacturers. More specifically, Nvidia-powered computers are expected both from Microsoft's Surface brand as well as other computer makers including Dell, and Microsoft and Nvidia will unveil the new PCs at the Computex trade show in Taiwan and Microsoft's Build developer conference in San Francisco.
Why developers should care: Microsoft is also expected to debut software aimed at enabling AI agents to perform tasks locally on Windows computers. That's the part the headlines are missing. Local agent execution rewrites the cost, latency, and privacy math for every product team building AI features into Windows apps. The strategic stakes are even bigger. If the first machines arrive with Surface, Dell, or other top-tier OEM backing, the Windows PC business may be entering its most consequential architectural contest since the original Copilot+ PC rollout.The official X accounts of Windows, Nvidia and chip design firm Arm all teased an upcoming announcement on Friday, announcing "A new era of PC," along with what appeared to be coordinates in Taiwan's capital Taipei.
If you're at Build in person, you'll be in the room where the Windows hardware story pivots. If you're remote, line up the keynote livestream.
Microsoft plans to stream the keynote and select Build 2026 sessions, and you can already register for free to livestream them. By registering today, you'll also get access to a session scheduler and on-demand keynote and sessions.Microsoft Build 2026 Day 2 is a live stream focused on developer tooling and sessions across GitHub, VS Code, Copilot, Windows, and Azure AI Foundry, including GitHub Copilot deep dives, VS Code announcements and demos, live coding with Microsoft and GitHub engineers, Windows sessions, Azure AI Foundry sessions, community speakers and creators, and select breakout sessions streamed live.
🔗 Register for the free Build 2026 livestream →
If you're building hosted agents, the through-line for this Build is unmistakable: agents are moving into production, and the teams that win are the ones who instrument, govern, and extend them properly from day one. Block off the keynote, the Foundry-on-Windows session, and at least one of the three workshops above - LAB540 first.
Bring your laptop, your hardest agent failure story, and a willingness to break things in the sandbox. We'll see you there.
Join Our Newsletter
