Visual Studio Code v1.116 Changelog

This release is all about making VS Code's AI agents more capable and easier to work with. If you use the AI chat to write code, run commands, or automate tasks, these updates give you more control, better performance, and tighter integration with your terminal.

What's new

• Copilot is built-in now — You can start using AI chat right away without installing the GitHub Copilot Chat extension separately.
• Agent Debug Logs — A panel that shows a timeline of everything the AI agent did during a chat. You can now view logs from past sessions too (they're saved on your computer), so you can look back and figure out what went wrong even after closing a chat.
• Thinking effort for Copilot CLI — Some AI models are "reasoning models" that think step-by-step before answering. You can now pick how much effort they put in (low, medium, high, etc.). Lower effort = faster responses; higher effort = more thorough answers. Non-reasoning models won't show this option.
• Terminal tools for agents — AI agents can now interact with any terminal you have open (like a running Python REPL or an interactive script), not just hidden background terminals they create themselves.
• Customizations welcome page — There's a new overview screen for all your agent customizations (instructions, skills, etc.). Not sure where to start? You can describe what you want in plain English and VS Code will draft the customizations for you.
• Tool confirmation carousel (Experimental) — When the AI wants to run several tools in a row, you now get a compact carousel to approve or reject them one by one instead of scrolling through a long conversation.
• VS Code Agents app (Insiders only) — A companion app designed for agent-first development got updates: reasoning level selection, automatic plan mode for planning tasks, better theming, and it now shows the Files tab by default in the Changes panel.

Improvements

• Smarter terminal input handling — The AI no longer makes an extra call to figure out if the terminal is waiting for input. This makes things faster and wastes fewer tokens (the chunks of text AI models process). The agent now handles input directly and asks you when it's unsure.
• Clearer progress messages — When the agent sends an answer to the terminal, the progress message now shows exactly which question it's answering (e.g., *Sending "my-project" to terminal (replying to: What is your project name?)*).
• Focus Terminal button — When the agent needs you to type something in the terminal (like a password), the confirmation carousel now has a button that jumps you straight to that terminal. If you just start typing, the carousel dismisses itself automatically.
• Background terminal notifications on by default — When an agent runs a command in a background terminal, it now automatically gets notified when the command finishes, times out, or needs input—no more polling to check.
• Diffs right in the chat — Code changes now show up directly in the conversation as diffs, so you don't have to switch to a separate view to see what the AI wants to change.
• Faster chat rendering — Responses appear faster thanks to reduced layout shifting and more efficient updates as the AI streams its answer.
• Faster message sending — Your message now shows up in the chat immediately, even if customizations are still loading in the background.
• Better subagent progress — The expanded view showing what sub-agents (smaller AI tasks spawned by the main agent) are doing is now more readable.

Fixes

• Rapid tool updates causing hangs — Fixed a bug where lots of tool updates happening at once caused the editor to briefly freeze.
• Chat send blocked by customizations — Fixed an issue where sending a chat message was delayed while customizations loaded; messages now appear instantly.

Breaking changes

No breaking changes in this release.