Starting today, GPT-5 is rolling out to all paid GitHub Copilot plans. GPT-5 is OpenAI's most capable model yet, bringing new advances in reasoning, coding, and chat.
🛠️ Enable more than 128 tools per chat request with new virtual tool grouping (experimental).
The model now automatically manages tool groups when you exceed the limit, so your request won't fail.
✅ Chat Checkpoints: create restore points in chat sessions to revert workspace changes and return to specific conversation states.
Perfect for multi-file editing when you want to experiment without fear.
⚙️ Setting: chat.checkpoints.enabled
We heard your frustrations with terminal hangs in chat – so we refactored the run in terminal and task tools for improved reliability and stability.
We also added output polling so the agent now waits for tasks and background terminals to complete before proceeding.
🔍 The #codebase tool now supports remote indexes for workspaces that are linked to Azure DevOps repos!
We are gradually rolling out support for this feature on the services side.
We revamped the tool picker experience. Now you can expand or collapse tool sets, MCP servers, and extensions with smooth sticky scrolling and enhanced icon rendering for better navigation.
⚡ Auto-start MCP servers without a manual refresh!
Configure automatic startup for new and updated servers, with trust dialogs for security.
⚙️ Setting: chat.mcp.autostart:newAndOutdated
We enhanced the terminal auto-approve settings with regex flags and improved logging, and we merged into a single setting: .terminal.autoApprove ⚙️chat.tools
🔔 We now leverage the OS native notification system to show a toast when user input is required but VS Code isn't in focus. Never miss important chat confirmation requests!
⚙️ Setting: chat.notifyWindowOnConfirmation
Git worktree support is here!
Check out multiple branches simultaneously, making it easy to test changes or work in parallel without switching contexts.
⚙️ Setting: git.detectWorktrees
Chats now have initial support for rendering mathematical equations in responses, powered by KaTeX.
This is a preview setting and can be enabled via the setting: chat.math.enabled ⚙️
🤖 When you start a coding agent session, you can now track your PRs with interactive cards right in chat.
There’s a new dedicated chat editor for managing coding agent sessions. Follow the progress of the coding agent, provide follow-up instructions, and jump between contexts seamlessly.
⚙️ Setting: chat.agentSessionsViewLocation
🔎 Join us for a release livestream next week to dive into this release with VS Code engineering team and other special guests!
The Prompt‑Driven Development series shows how to turn prompts into real outcomes: refactoring existing codebases, fixing a chat app, querying docs with MCP, building responsive UIs, improving documentation, and generating tests.
🧵 Start here 👇
1️⃣ Introduction to Prompt‑Driven Development
Before diving into advanced workflows, this video breaks down the foundations: what AI‑assisted coding really is, how PDD differs from vibe coding and prompt engineering, and the core practices that make your prompts reliable.
2️⃣ Refactor an Existing Codebase using Prompt‑Driven Development
How to refactor a FastAPI + Cosmos DB project with Copilot using clear prompts, Agent Mode, and step‑by‑step reviews.
With the latest @code release, the Language Model Chat Provider extension API is finalized - what does that mean for you?
Models can now be contributed via extensions, so with a single install, you get more model choice via Bring Your Own Key.
Here are some extensions you can install now… 🧵
The AI Toolkit extension enables you to explore and evaluate models.
And now you have access to its provided models directly in @code, whether it's a custom model you've tuned in Azure AI Foundry, a local model via Foundry Local, or any of the models in GitHub Models. aka.ms/AIToolkit
@huggingface's Inference Providers give developers access to hundreds of LLMs, built for high availability and low latency.
With the Hugging Face Provider extension, you can use models like Kimi K2, DeepSeek V3.1, GLM 4.5 directly in @code aka.ms/VSCode/Hugging…
Big updates in the latest @code release:
✨ Extend agent mode (now available in VS Code Stable!) with MCP tools
✨ A unified chat experience
✨ Use your own API key to access more language models
...and so much more