The latest Azure Logic Apps (Standard) extension for Visual Studio Code (VS Code) now supports .NET 10 custom code. You can create new .NET 10 local-function projects and migrate your existing .NET 8 projects without turning them into separate Azure Functions apps.

Check your prerequisites

• Install the Azure Logic Apps (Standard) extension for VS Code, version 5.1013.2.
• Use Logic Apps Standard extension bundle 1.194.31.

After the extension's dependency setup completes, the .NET 10 software development kit (SDK) and runtime should be installed in its managed dependencies folder. On Windows, the default .NET location is:

%USERPROFILE%\.azurelogicapps\dependencies\DotNetSDK

Check that installation, rather than relying on the dotnet command elsewhere on your PATH. If you've customized the extension's dependency location, check the configured location instead.

New projects target .NET 10

When you create a new custom-code project through the extension, .NET 10 is now the only supported target. The .NET 8 option has been removed for new projects.

Existing .NET 8 projects still need a configuration migration before they target .NET 10. Updating the extension isn't the same as retargeting your project. Unchanged project will continue to run on .NET 8 until the configuration changes below are applied.

If you use the C# code action

If your workflow uses the C# code action, add the runtime setting to your Logic Apps project's settings - either in a local project or on an Azure application. Set LOGIC_APPS_CUSTOMCODE_DOTNETVERSION to net10.0.

Migrate an existing .NET 8 project

You don't need to start with a new workspace. Back up your workspace, including local configuration, before making the changes below. Paths are relative to each project's folder.

Changes in the custom-code project

Those package versions describe the configuration we tested, not a universal minimum dependency set. Resolve the dependencies your project actually needs rather than copying every package from a sample.

A sample project file will look like this

<Project Sdk="Microsoft.NET.Sdk"> <PropertyGroup> <IsPackable>false</IsPackable> <TargetFramework>net10.0</TargetFramework> <AzureFunctionsVersion>v4</AzureFunctionsVersion> <OutputType>Library</OutputType> <PlatformTarget>AnyCPU</PlatformTarget> <LogicAppFolderToPublish>$(MSBuildProjectDirectory)\..\{your logic app target folder}</LogicAppFolderToPublish> <CopyToOutputDirectory>Always</CopyToOutputDirectory> <SelfContained>false</SelfContained> <CopyLocalLockFileAssemblies>true</CopyLocalLockFileAssemblies> </PropertyGroup> <ItemGroup> <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.Abstractions" Version="1.3.0"/> <PackageReference Include="Microsoft.Azure.Functions.Worker" Version="2.51.0"/> <PackageReference Include="Microsoft.Azure.Functions.Worker.Sdk" Version="2.0.7"/> <PackageReference Include="Microsoft.Azure.Workflows.Webjobs.Sdk" Version="1.4.0"/> <PackageReference Include="Microsoft.Extensions.Logging.Abstractions" Version="10.0.0"/> <PackageReference Include="Microsoft.Extensions.Logging" Version="10.0.0"/> </ItemGroup> <Target Name="TriggerPublishOnBuild" AfterTargets="Build"> <CallTarget Targets="Publish"/> </Target> </Project>

Changes in the Logic Apps project

Verify the migration

Save all files, restore, and build. Confirm publication under lib\custom\net10.0 in the Logic Apps folder, then run the workflow that calls your edited function and verify its breakpoint.

Use the ready-made migration skill

We've prepared the logicapps-custom-code-net10-migration skill to guide a coding agent through the same process. It inspects your workspace, plans configuration changes before applying them, and guides build and publication verification. It preserves your function and workflow implementations rather than rewriting them.

The skill is available in the custom-code-net10 repository, while we work toward contributing it to Awesome Copilot.

With GitHub CLI (command-line interface) version 2.90.0 or later, inspect the skill and install it for Copilot across projects. These skill commands are currently in preview.

gh skill preview wsilveiranz/custom-code-net10 logicapps-custom-code-net10-migration

gh skill install wsilveiranz/custom-code-net10 logicapps-custom-code-net10-migration --agent github-copilot --scope user

Installation makes the skill available to Copilot; it doesn't run the migration. If you prefer a download, the v0.1.0 release also includes a ZIP of the complete skill folder.

After installation, you can ask your agent:

Migrate this Logic Apps Standard .NET 8 custom-code workspace to .NET 10 using the logicapps-custom-code-net10-migration skill. Preserve my workflows, back up the project first, and include coordinated VS Code debugging.

Keep the host and deployment settings in view

The custom-code target framework isn't the Logic Apps host runtime. Keep the existing host settings, including FUNCTIONS_WORKER_RUNTIME=dotnet. Don't switch the host to dotnet-isolated just because the custom code targets .NET 10.

Preserving your existing code is the migration's starting point, not a guarantee that it will build and run unchanged. Use the build and an executed local function to check the result.

For Azure execution, the deployed Logic Apps application settings must include both LOGIC_APPS_CUSTOMCODE_DOTNETVERSION=net10.0 and FUNCTIONS_INPROC_NET8_ENABLED=1. The .NET 8 host flag is still required even though your custom code targets .NET 10.

When you deploy through the Logic Apps extension in VS Code, these settings are copied to the Azure app. If you deploy through a pipeline, make sure your deployment configuration explicitly includes both application settings. A local settings file alone doesn't configure the Azure app, so verify the deployed runtime configuration after deployment.

The migration skill doesn't change Azure resources as a side effect.

Try the updated extension with one of your custom-code projects. If migration or debugging needs a step we've missed, share the configuration and the symptom in the comments, without credentials or private data. I want to hear where the build or debugging experience still gets in your way.

 

Where building brings people together

What happens when people who love building things find themselves in the same place? At GitHub Universe 2026, held October 28–29, Microsoft MVPs, several of them also GitHub Stars, took the stage, shared ideas, and swapped stories. Their reflections go beyond new tools to show why sharing knowledge matters: a handmade nameplate, hallway conversations, and practical lessons on AI context, security, and policy turned technical exchanges into connections - and offered ideas to bring home to their communities.

GitHub Universe Speaker Nick Taylor (MVP)

Where Microsoft MVPs and GitHub Stars come together

Several Microsoft MVPs are also GitHub Stars, and GitHub Universe is where the two communities come together. The GitHub Stars program recognizes community leaders who inspire, educate, and shape communities, giving them a platform to showcase their work, reach more people, and help shape the future of GitHub. Perks include early access to GitHub features, invitations to GitHub events, connections to internal GitHub Team members and more. MVPs Kent C. Dodds, Huan Li, and Nick Taylor who spoke at GitHub Universe are GitHub Stars.

For Huan, the moment that captured the spirit of GitHub Universe was sitting on the floor beside the GitHub Star with his name on it - and yes, he needed another photo with it.

“I see it every year, and somehow it still makes me smile.” - MVP and RD Huan Li

To learn more about the GitHub Stars, visit stars.github.com.

GitHub Universe Attendees (left to right) Esteban Garcia (RD & MVP) andAnisha Pindoria (GitHub)

Conversations, creativity, and connection

Ask a Microsoft MVP what made GitHub Universe stand out, and the answer is rarely a single session. MVP Sam Basu says the event “always feels like a different type of tech conference,” and because it is GitHub, “it is truly polyglot.” It is a chance to “rub shoulders with developers across variety of tech stacks, from various corners of the world.”

“The vibe is truly unparalleled with focus on building, collaborating & having fun learning.” - MVP Sam Basu

 

GitHub Universe Attendees, (left to right) David Fowler(Microsoft) and Sam Basu (MVP)

For many MVPs, that focus on building was literal. MVP and Regional Director Huan Li spent almost an hour putting hundreds of tiny pieces together to build his own nameplate, and it was “worth every minute.” His takeaway: “even when I take a break from building software, I still want to build something!”

MVP and Regional Director Huan Li built a custom nameplate at GitHub Universe

MVP Dylan McCarthy saw the same enthusiasm in the makerspace, where developers lined up to build custom keyboards and GitHub lamps. “They were so popular that supplies ran out before I even managed to make one myself!” he says.

Some of the best moments happened between sessions. MVP and Regional Director Esteban Garcia points to “the conversations in between everything else,” talking with people about “what they were building, what was working, and what they were still trying to figure out.” MVP Josh Garverick agrees. His favorite part of attending is “getting to talk with my friends in the community, Hubbers, and new contacts.”

“This continues to foster my appreciation for the community and fuels me to do as much as I can to contribute.” - MVP Josh Garverick

Ideas you can put into practice

MVPs also brought their expertise to the stage. In “Shut the context window! You’re letting the tokens out,” MVP Dylan McCarthy explored how Copilot uses context and tokens. His advice: “moving away from large, monolithic GitHub Copilot instruction files and instead using smaller, targeted instructions with applyTo patterns.”

“It’s not simply about reducing token consumption, but about making sure the information we’re providing is relevant, useful, and helps the model deliver better results.” - MVP Dylan McCarthy

Zooming out to a bigger question, MVP Kent C. Dodds took the Main Stage to ask, “What’s worth building when AI can build anything?” Joined by Dr. Sue Black of Durham University, he made the case for building deliberately rather than reflexively, exploring responsible software, the daily reality of shipping, choosing between tools, and maintaining what’s already built. The conversation continued at a Session After Party in GitHub Central, an informal Q&A for attendees’ questions. 

Watch Kent and Sue’s Main Stage conversation to bring a sharper set of questions to your own work.

Building deliberately also means deciding what AI agents are allowed to do. In “Open PRs, don’t merge them: Fine-grained authorization for hosted MCP servers,” MVP Nick Taylor showed an identity-aware proxy that adds per-identity authorization in front of a hosted MCP server, with no changes upstream. In a live demo, Copilot opened a pull request and left the merge to a reviewer, enforced per user. Nick also pointed to Martin Woodward’s closing keynote, a reminder that AI is a tool and people still bring the ideas. Watch Nick’s session to see the live demo and the pattern for your own MCP servers.

“You can improve your security posture immediately with an identity-aware proxy, whether that be putting a legacy app behind it or building out an MCP gateway with it.” - MVP Nick Taylor

Policy matters just as much. What happens when a prototype starts handling real customer payments but its repository still carries its launch-day merge requirements? In “When a prototype becomes critical, does your GitHub policy know?,” MVP Baruch Sadogursky showed how he syncs ownership and business criticality from Port into GitHub as custom properties, so teams can find repositories by owning team and point a GitHub ruleset at critical services. The pattern is repeatable: Port supplies the context, and GitHub turns it into enforced policy. Watch Baruch’s session to see the gap close live.

Context ties these sessions together. Sam Basu spoke about bringing context to AI tools through Skills, MCP tools, and custom agents, and his “full room with engaged audience” showed how much this is on people’s minds. Honest conversations about balancing enterprises’ privacy and cost concerns with AI’s benefits were among the best things he took from GitHub Universe. Esteban Garcia offers a simple way to start:

“Build something. Try it with your team. Pay attention to where it works and where it falls apart, and then share what you learn.”

- MVP & RD Esteban Garcia

Keep building. Keep sharing.

Curious about GitHub Universe? Explore the recorded sessions or browse the session catalog, pick an idea that interests you, and invite someone to try it with you. Then share what you learn with your own community. Whether you are a developer, engineer, IT professional, or simply curious about technology, there is a conversation you can contribute to. What could you build, or help someone else learn, before the next conference? 

Want to learn more about the MVP Program?

To find an MVP and learn more about the Microsoft MVP Program, visit the MVP Communities website and follow program updates on LinkedIn or through #MVPBuzz.

Mixed Reality Link now lets you stream your Windows desktop to your headset over USB. This wired option gives your Meta Quest 3 or 3S headset a direct connection to your PC when Wi-Fi is congested or unavailable.

Stream without Wi-Fi

Wi-Fi is a great way to work untethered, but not every network offers a high-quality connection between your PC and headset. USB gives you another option when coverage is poor, the network is congested, or public Wi-Fi prevents devices from communicating with each other. USB is particularly useful when traveling: once set up, you can use your PC in mixed reality even without internet access.

Simplified pairing

Pairing a headset for the first time is also simpler over USB. There’s no QR code to scan or PIN to enter, making it faster to establish a secure connection between your PC and headset.

Get started

You’ll need a Windows 11 PC running Mixed Reality Link version 26.10120.9280.1 or newer, a Meta Quest 3 or 3S headset, and a USB data cable. Meta is rolling out USB support over the next week, so your headset may not have it yet, even if you have the latest version of Mixed Reality Link.

Once the experience is available on your headset, follow these steps to get connected:

1. Install or update Mixed Reality Link on your PC.
2. On your Quest, go to Settings → Link and enable “USB connection for apps”.
3. Connect your headset to your PC using a USB cable that supports data.
4. Open Virtual Display on your headset, then follow the prompts to pair with your PC or connect to a previously paired PC.

For more information, visit our support page. We’d love to hear how USB works for you. Share your feedback and help us improve Mixed Reality Link.

Video workloads do not all start with the same input or aim for the same result. A creative team might explore several interpretations of a written brief. A retailer might turn an approved product image into a short motion concept. An application developer might combine image generation and video generation into a workflow that takes an idea from description to frame to scene. Each scenario brings different requirements for visual quality, motion, creative control, turnaround time, and cost. All of this means that a multi-model solution is necessary to realize a full vision.

Today, Microsoft and fal are helping developers find the right tools for their workloads. We are excited to announce that Kling, LTX, and MiniMax video models are rolling out to in public preview in Microsoft Foundry through our new partnership with fal. Available as a first-party Azure service, these text-to-video and image-to-video options give creators and developers more ways to utilize video generation with the experiences they want to build.

The partnership expands the choices available in Microsoft Foundry while bringing video into the broader platform developers use to build AI applications. The goal is to let developers pick the right model for each step of an experience, then connect those steps into a working application.

“fal has seen incredible adoption this year alone with generative media across all industries from advertising to marketing to gaming. Partnering with Microsoft allows both of us to truly enable the wide creative ecosystem and ensure that all AI solutions that lean on generative media, becomes a multi-model one.” - Burkay Gur, Cofounder

Bringing fal’s generative media focus to enterprise builders

fal is the full stack for generative media, built for creators and developers, with APIs for over 1500 models across image, video, audio, 3D, and generative worlds. As a truly encapsulating platform, fal provides inference, training, and infrastructure to run generative media workloads in production, while also providing innovative post trained models that pushes the boundaries on speed and quality.

That specialization matters because generative media is both a creative and an engineering workload. A compelling demo is only the start. Production applications also need to handle real traffic, scale with demand, and balance output quality, latency, and cost.

Through this partnership, developers will be able to access a selected set of video models in Microsoft Foundry, served by fal and sourced from the companies that build them. The initial portfolio gives teams a choice of models for different creative requirements instead of relying on one model for every task, with more models planned over time.

A portfolio for different video experiences

The initial portfolio spans Kling v3 Pro, LTX-2.5 Fast, LTX-2.5 Pro, MiniMax H3 Max, and MiniMax H3 Max Turbo. These models give developers different starting points for generating video from text and animating existing images. A team rapidly exploring creative concepts may prioritize speed and cost, while a team producing fewer, higher-fidelity assets may weigh output quality more heavily. Whether an application starts from text or an existing image, and what it needs in output quality and cost, also shapes the choice.

Kling: From a still image to a directed scene

Kling v3 Pro combines cinematic visuals and fluid motion with native audio generation and custom element support. Its image-to-video capabilities give developers control over how a starting image becomes a moving scene, a fit for applications that prioritize visual direction and motion.

LTX: One model for fast drafts, one for final output 

For rapid iteration and previews, LTX-2.5 Fast is the speed-optimized option; for high-fidelity final output, LTX-2.5 Pro is the quality-optimized option. Both support text-to-video and image-to-video generation with synchronized audio.

MiniMax: Another approach to video generation

MiniMax H3 Max and MiniMax H3 Max Turbo are post-trained variants of MiniMax H3 focused on prompt adherence, aesthetics, and inference performance. For developers, they provide additional options to evaluate against an application’s creative brief, quality requirements, and performance needs.

“When building with video models, the details matter: how closely the output follows your prompt, how well it preserves your reference images/videos/audios and how quickly you can try another idea. Kling gives developers control over detailed reference to video generation. LTX offers fast iteration and high-quality output through its Fast and Pro variants. H3 Max brings fal's post-training and inference optimizations to improve quality and speed. Having these models available through Microsoft Foundry gives developers more flexibility to choose what works best for the application they are building." - Gökay Aydoğan, Creative Engineer

Built for enterprise video workloads

Bringing video generation into an enterprise application requires more than choosing the right model. Teams need a platform where they can evaluate different options, build with the models that fit their requirements, and operate those experiences alongside the rest of their applications.

The initial portfolio of video models is hosted in the US, and Microsoft Foundry is the first provider to offer a US-hosted Kling endpoint. Key considerations for enterprise teams:

• Data residency: For organizations with data residency requirements, hosting location is one factor to weigh. Review each model's documented data processing and storage terms to confirm it meets your requirements. 
• Model evaluation: Teams can assess quality, performance, and cost, then connect video generation with the other models and capabilities their applications need. 
• Production readiness: As they move toward production, they can use the evaluation, optimization, governance, security, observability, and infrastructure capabilities supported for their selected offering.

Pricing

Video generation pricing varies by model and output resolution, giving teams options to balance creative requirements with their application’s budget. The table below shows pricing in USD per minute of generated video for LTX and MiniMax models, making it easier to compare costs across the available options.

Pricing will be updated and listed on our Azure pricing page within the coming weeks. 

Get started

Ready to start building? Join us at Ignite for BRK256: Video AI at Scale with fal on Microsoft Foundry. Naomi Moneypenny and Alexander Hughes from Microsoft will be joined by Burkay Gur and Lovis Odin from fal on stage to explore this partnership on Thursday, November 19 at 11:00 AM - 11:45 AM CT. Register today!

Explore Kling v3 Pro, LTX-2.5 Fast, LTX-2.5 Pro, MiniMax H3 Max, and MiniMax H3 Max Turbo in Microsoft Foundry and find the model that fits your video experience. 

• Explore the models in Microsoft Foundry 
• Learn more about fal 
• Read the quick start guide to using fal models with ComfyUI here 
• If you are interested in testing out new models, check the Microsoft Foundry Model Releases Repository

 

 

The hardest work for an AI agent rarely fits into a single prompt. A coding agent may need to follow a change across a large repository. A research agent may need to work through lengthy documents, charts, and images before it can give a useful answer. In both cases, the challenge is not just generating a response, but keeping track of the work as it unfolds 

Today, Moonshot AI’s Kimi K3.1 Pro is coming to Microsoft Foundry. It is an open-weight model built for agentic work, with text and image understanding and a context window of up to 1 million tokens. That combination makes it an interesting choice for developers building coding, research, and document-heavy applications. 

 

Where Kimi K3.1 Pro stands out 

Kimi K3.1 Pro is the first model in Moonshot AI’s K3.1 series. It builds on Kimi K3 with improvements across coding, research, and other multi-step tasks. Its large context window gives applications room to bring more relevant material into a task, while its multimodal capabilities let the model work with images alongside text. 

That opens up practical possibilities: 

• Coding across a larger body of work. Give an agent the context it needs to investigate a change, understand related files, and work through an engineering task. 

• Research grounded in source material. Bring lengthy reports and supporting material into a workflow so the model can examine the evidence before producing a synthesis. 

• Work that combines text and visuals. Build applications that need to interpret images, charts, or document pages as part of a broader task. 

A large context window is useful, but it is not a shortcut to a good application. What matters is whether the model can use that context effectively, complete the task, and do so at a cost and speed that work for the customer. 

 

Choose based on the work your users need done 

Moonshot AI’s benchmarks have strong results for Kimi K3.1 Pro across coding, agentic, and multimodal evaluations. Those benchmarks are a useful starting point, but they do not answer the question every developer ultimately has: Will this model perform well in my application? 

The best way to find out is to test it on representative tasks. Compare the quality of the completed work, not just the first response. Look at how reliably the agent handles long inputs, where it needs tools, how it performs with visual material, and the cost per successful task. 

That is the value of bringing another capable model into Microsoft Foundry. Developers can evaluate Kimi K3.1 Pro alongside other models and make a choice grounded in their own workload, rather than assuming one model is right for every step. 

 

Pricing table  

 

Get started in Microsoft Foundry 

Explore Kimi K3.1 Pro in the Microsoft Foundry model catalog..  Start with a real coding, research, or document workflow, establish a baseline with the model you use today, and see where Kimi K3.1 Pro makes a meaningful difference. The opportunity is not simply to fit more into a prompt. It is to build agents that can stay with a complex task long enough to deliver a better result. 

 We understand from customers how important a variety of open weight models are for their specific workloads and needs. Today’s launch of Kimi K3.1 Pro on Microsoft Foundry is the latest in our ongoing investment in the open weight model ecosystem. Microsoft Foundry features a wide range of open-weight models from providers including NVIDIA, OpenAI, DeepSeek, MiniMax, Mistral AI, Moonshot AI, and Qwen.  Even more models are available through partners, such as Fireworks on Foundry.