You Don’t Need a GitHub Copilot Subscription to Use VS Code AI Features | by Jeff Flynt | Jun, 2026 | MediumSitemapOpen in appSign up<br>Sign in

Medium Logo

Get app<br>Write

Search

Sign up<br>Sign in

You Don’t Need a GitHub Copilot Subscription to Use VS Code AI Features

Jeff Flynt

4 min read·<br>Just now

Listen

Share

Press enter or click to view image in full size

Short answer: no.<br>You need the GitHub Copilot extension and a GitHub account, but you don’t need a paid Copilot subscription.<br>In fact, these models are coming from Tenure, not GitHub:

Native VS Code AI features using a custom language model provider.Most developers assume VS Code AI and GitHub Copilot are inseparable.<br>They aren’t.<br>Everyone assumes these are the same thing<br>If you’ve been using VS Code for a while, you probably have a mental model that looks something like this:<br>VS Code AI = GitHub Copilot = Monthly subscriptionThat’s a reasonable assumption. GitHub ships the default experience, it’s the thing everyone talks about, and for most developers it’s the only AI experience they’ve seen inside VS Code. So it feels like one thing.<br>It isn’t.<br>VS Code AI and GitHub Copilot are actually separate things<br>Here’s how it works under the hood.<br>The Copilot extension provides the interface: chat, inline completions, code actions, agent workflows. That part is GitHub’s. But the actual language model behind those experiences doesn’t come from the Copilot extension directly. It comes through something called VS Code’s Language Model APIs, which sit between the interface and whatever is actually generating responses.<br>VS Code<br>Copilot UI<br>Language Model APIs<br>Provider<br>ModelThat middle layer, the Language Model APIs, is where things get interesting. It’s an open API. Any extension can register itself as a provider. GitHub’s provider is just one implementation of it.<br>Most developers never need to think about this distinction because they’re using GitHub’s provider and GitHub’s models and everything is bundled together. But the architecture was always more open than the default experience suggests.<br>Which means you can bring your own models<br>Once you understand that the provider slot is swappable, the possibilities are pretty straightforward. A custom provider could route to:<br>OpenAI<br>Anthropic<br>Google Gemini<br>Azure OpenAI<br>AWS Bedrock<br>Local models running on your own hardware<br>Anything else that speaks a compatible API<br>And crucially, the developer experience stays exactly the same. You’re still using VS Code’s chat panel. Still getting inline completions as you type. Still triggering code actions from the right-click menu.<br>The interface doesn’t change because the interface is Copilot’s. Only the intelligence behind it changes.

Why we built Tenure this way<br>Most AI tools ask developers to change their workflow. New interface, new panel, new place to switch context. We wanted the opposite.<br>When we built Tenure, the question was: how do we make organizational AI memory feel invisible? The answer was to meet developers where they already are, which is inside VS Code, using the tools already built into the editor.<br>So Tenure registers as a native language model provider. When you open VS Code and sign in with GitHub, Tenure’s models show up in the Copilot Chat model picker alongside everything else. There’s no secondary panel. No browser tab. No separate app.<br>VS Code<br>Tenure LM Provider<br>Automatic Project Scope<br>Organizational Memory<br>Your ModelsThe other piece is project scope. VS Code already knows which workspace you have open. Tenure watches for file switches and synchronizes that information to the proxy before your first message is sent, so every request is automatically scoped to the right project. You don’t have to tell it where you are. It already knows.<br>Press enter or click to view image in full size

Architectural decisions and team knowledge are injected before requests reach the model.The result is that developers keep using the AI features built into VS Code while their organization keeps control over which models are running, what context is available, and where the memory lives.<br>The part that surprises people<br>When we explain this to developers, the reaction is usually some version of “wait, seriously?”<br>Because the implication is real: you can have the full native VS Code AI experience, chat, completions, code actions, everything, running on your own models, with your own context, and you’re not paying GitHub a subscription fee for it.<br>What you do need:<br>A GitHub account (free)<br>The GitHub Copilot extension installed (free)<br>That’s it. The extension needs to be there because it provides the interface. But an active paid Copilot plan is not required. The free tier is enough. Tenure acts as the provider, so the model requests never touch GitHub’s infrastructure.<br>The architecture was always there. VS Code already separates the interface from the model. GitHub ships the default provider, but it isn’t the only provider. Which means you can keep the native VS Code...