The best AI in Visual Studio is the AI you can bring with you. Developers don’t work in a single-model world anymore. You might reach for one model for everyday coding, another for a domain-specific problem, and an approved model for work that must stay within your organization’s governance boundaries. So, the question we hear most often isn’t which model is best. It’s, can I use mine?
We’re bringing Bring Your Own Model (BYOM) in Preview to Visual Studio, making it available and enabled by default across Community, Professional, and Enterprise SKUs. This Preview focuses on the developer experience available today. Enterprise management capabilities, including admin controls to disable BYOM as well as centralized model configuration and management for Professional and Enterprise SKUs are coming soon.
From individual developers experimenting with new models to organizations standardizing approved deployments, BYOM gives you the flexibility to use AI in Visual Studio your way. With BYOM, you can connect to a Microsoft Foundry deployment, or use your own key from another supported provider, in Agent Mode, whether or not you’re signed in to GitHub.
This is an early Preview and we’d appreciate feedback on the developer experience. Before you dive in, watch the short demo below to see the capability in action. Then try it yourself by following the steps in the Getting started section below and let us know what’s working, what’s missing, and where you’d like us to invest next. Share your feedback in the Developer Community forum and/or fill out this short survey to help shape the future of BYOM in Visual Studio.
Getting Started
Important: The BYOM capability is evolving alongside Visual Studio’s new Agent (Preview), which is built on the new GitHub Copilot SDK-powered harness. As part of this transition, the previous BYOM experience available in the earlier Ask and Agent modes is no longer supported. To use BYOM, install the latest Visual Studio Insiders release and use the new Agent (Preview) experience. If you had previously added BYOM models, they will need to be re-added with this update.
BYOM is available in Preview within the Visual Studio 18.10 Insiders Release.
To try it:
- Install Visual Studio 18.10 Insiders Release
- Within Chat, open the model picker and choose ‘Add a model’ or ‘Manage models’ option (this will work for both GitHub signed-in and signed-out flows)
- Connect to the provider of your choice, by clicking on ‘Add model provider’
- Start working in Agent (Preview) Mode
Supported providers for the Preview: Microsoft Foundry, OpenAI, Anthropic, Ollama with custom URL support for OpenAI & Ollama.
We also want to be candid about what Preview means. Not every model supports every Agent Mode capability today, and we’re not attempting to certify every model-and-provider combination in this release. Where a feature depends on a capability a model doesn’t offer, we aim to fail gracefully.
Why developers asked for BYOM
Over the past several months, we’ve spent a lot of time with enterprise developers, IT administrators, and security leads, the people who decide whether an AI coding tool gets approved, restricted, or blocked. One theme came through clearly: teams increasingly work in multi-model environments, and they want Visual Studio to meet them there.
A few motivations came up repeatedly as reflected in this suggestion ticket:
- Organizations have already invested in approved models, often deployed behind private endpoints, and want developers to use them with Copilot in Visual Studio without sending code or prompts to another provider.
- In regulated industries, including financial services, healthcare, the public sector, and defense, tools must meet organizational requirements for identity, data flow, and administrative control to be usable.
- Cost and performance matter. Teams want the flexibility to choose models that meet their budget and latency requirements.
BYOM gives developers and organizations more choice in how they use AI in Visual Studio. Use GitHub Copilot models, access organization-approved deployments from Microsoft Foundry or connect your own provider. Developers stay in the same Visual Studio experience, while organizations retain the governance, security, and compliance controls they’ve already established. The result is more flexibility to use the models that best fit your workflow, budget, and organizational requirements.
What we’re working on next
This Preview is the first step toward Bring your Own Model in Visual Studio. Next, we’re focused on making BYOM easier to adopt and govern across organizations through an ADMX policy to disable it for Professional and Enterprise users that’s coming soon as part of the 18.10 GA release, centralized model configuration and management, more granular model controls such as configuring context size, tool calling, and model thinking effort, broader provider and model support, smarter compatibility detection and deeper Microsoft Foundry integration. These capabilities are not included in the current release, and your feedback will help us prioritize what comes next.
Try BYOM in Visual Studio 18.10 Insiders today and share what’s working, what’s missing, and what you’d like us to prioritize next through the Developer Community forum or our short survey. Your feedback will directly help shape the future of AI in Visual Studio. Happy Coding!
0 comments
Be the first to start the discussion.