Visual Studio Code v1.122 Changelog
What changed in VS Code v1.122 — explained in plain English.
- AI coding assistants can now handle massive amounts of code at once and work completely offline.
- GitHub Copilot switched to a "pay for what you use" billing model, so you can now see how much your AI questions cost.
- You can test how websites look on different devices right inside the editor.
- It's easier to report bugs to VS Code with built-in screen recordings.
Want a heads-up when Visual Studio Code updates?
We email you the moment a new Visual Studio Code version drops — plus plain-English release notes like these. No spam.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
This update is all about supercharging VS Code's AI features. It gives developers way more flexibility to use AI tools offline, keeps track of how much those AI tools cost, and makes it easier to test web apps, making it a big deal for anyone building software in restricted environments or just trying to manage their budget.
What's new
- Massive AI memory (1M context window): AI assistants like Claude and GPT can now "remember" up to a million tokens (pieces of words) at once. This means you can feed the AI huge chunks of your code or have super long conversations without it forgetting the beginning.
- Offline AI (Air-gapped BYOK): "BYOK" stands for Bring Your Own Key, meaning you use your own personal AI account instead of a built-in one. You can now use your own AI models without even logging into GitHub. This is awesome for highly secure, offline computers or if you just want to use local AI models like Ollama.
- Device testing in the browser: You can now test how your website looks on different screens (like phones or tablets) directly inside VS Code's built-in browser.
- Better bug reporting: If VS Code messes up, you can now send a bug report with screenshots and video recordings to show the developers exactly what happened.
- A new window just for AI Agents: There's a new "Agents window" designed to help you manage your AI assistants across different projects. When you hover over an AI session, you can quickly see what files it changed.
- Custom AI connections: You can now connect VS Code to your own private or company AI servers (called a Custom Endpoint) in the regular, stable version of the app.
Improvements
- Pay-as-you-go AI billing: GitHub Copilot now uses "usage-based billing." Instead of a flat fee, you spend AI credits based on how much text you send and receive.
- Cost tracking: The Copilot dashboard now shows how many AI credits you've used. Plus, when you pick an AI model, it shows you the price tag so you can choose a cheaper model for simple tasks.
- Safer AI commands (Sandboxing): When an AI agent runs computer commands for you, it used to try them in a "sandbox" (a safe, isolated space) even if you had approvals turned off. That was confusing, so now it only uses the sandbox when it actually helps keep your computer safe.
- Better AI monitoring: If a company wants to track what its AI agents are doing, the AI now sends better, more detailed logs to monitoring tools.
Fixes
- Missing AI features offline: When using your own offline AI models, small background tasks (like asking the AI to name your chat or write a commit message) used to break because they couldn't find the default AI. Now, a helpful pop-up will ask you to assign your offline AI to handle these small jobs so everything works smoothly.
Breaking changes
No breaking changes in this release.
This is our plain-language summary. Read the complete, official notes on the Visual Studio Code official changelog ↗.
Google Calendar