You open a new project someone sent you. VS Code immediately interrupts you with a giant trust dialog before you can even look at the code. You click trust, realize you maybe shouldn't have, and now you're quietly hoping nothing weird runs. Sound familiar? š
VS Code 1.126 just landed on June 24, 2026, and it brings some genuinely useful upgrades ā not just polish for the sake of polish, but real quality-of-life wins. We're talking AI cost transparency, running multiple agent chats side by side, and a smarter way to handle untrusted folders.
Let's walk through what's new and why it actually matters for your daily workflow.
What Is VS Code 1.126?
VS Code 1.126 is the latest monthly release of Visual Studio Code ā Microsoft's wildly popular, free, open-source code editor. Every month, the team ships a new version with improvements ranging from subtle UI tweaks to major feature drops.
This release focuses heavily on Copilot chat and agent workflows, making AI assistance in VS Code more manageable, more readable, and less of a mystery when it comes to cost. It also quietly improves how you handle folders you don't fully trust yet.
If you use VS Code daily (and let's be honest, most of us do š), there's something in here for you.
Why This Release Matters
AI tools inside editors are no longer a novelty ā they're becoming part of the actual development loop. But with that power comes a new kind of headache: you don't always know how much your AI interactions are costing you, or how to manage multiple threads of work at once.
This release tackles both of those problems head-on. It also quietly fixes one of those small-but-annoying security UX issues that has bugged developers for years. Whether you're a solo developer, part of a team, or just starting out, these changes make VS Code feel sharper and more intentional.
New Features and What They Actually Do
š° Session-Level Cost Information
This is a big one for anyone using Copilot with credit-based plans.
Previously, you could only see the cost of an individual chat turn. Now, VS Code shows you the total cost for an entire chat session ā not just one message, but the whole conversation combined. A session-level info popover also shows context window token usage across the whole session.
Why this matters: Imagine you're debugging a tricky async issue and you've sent 30 messages back and forth. Before this, you'd have no idea how many credits that conversation consumed until you checked your billing dashboard. Now, the information is right there in the chat UI.
It's like finally getting a running total at the grocery store instead of being surprised at checkout.
š§ Unified Model Customization Picker
VS Code used to give you two separate dropdowns to configure a language model ā one for context size and another for reasoning (thinking) effort. That was fine, but switching between them felt unnecessarily scattered.
Now, both settings live inside a single model customization picker. One place, two controls, done.
The model hover tooltip also got a cleanup. It now shows a concise one-word descriptor of the model's capabilities (instead of a wall of text), plus deep link buttons that take you directly to the settings you need.
Why this matters: When you're mid-task and want to quickly switch a model to "think harder" or reduce context size to save credits, the fewer clicks the better. This makes model tuning feel lighter.
šŖ Multiple Chats in One Agent Session
The Agents window (still in Preview) just got a significant upgrade.
Before, you could only have one chat running inside an agent host session. Now you can open multiple chats side by side within the same session. These chats share the same working context, which means:
- Start a chat implementing a new feature
- While it's working, open a second chat in the same session to write tests or review changes
- Both run at the same time
- Each chat keeps its own separate conversation thread
You can rename each chat tab so you don't lose track of what's what. And if you close VS Code and come back later, all your chats are persisted and restored ā not just the first one.
Why this matters: This is genuinely useful for complex, multi-step work. It's like having two terminals open side by side, except for your AI conversations. No more "wait, where was that other chat I had about the database schema?"
š¬ Agentic Code Feedback with Comments
If you use the Agents window for code review workflows, this one is for you.
Comments you leave on generated code are now stored on the agent host ā not locally in your browser session. This means:
- The agent can read and respond to your comments using server-side tools
- Comments persist even if you disconnect the client
- The agent itself can add inline comments using the
addCommenttool
When you run a review skill like /code-review, the agent reviews your code and drops inline comments that you can accept or delete before sending them back to the agent to address. Pull request review comments work the same way ā the agent can request permission to view PR comments and then address them.
Why this matters: This closes the loop between you and the agent during reviews. Instead of copy-pasting feedback, you're annotating directly and the agent picks it up.
š Safer Folder Browsing with Restricted Mode
This is the one that fixes that annoying trust dialog problem.
Previously, when you opened a new or unfamiliar folder in VS Code, it immediately hit you with a dialog asking whether you trust the folder ā before you could even look at the code. Rushing past that dialog and clicking "Yes" without thinking was easy, and trusting code you haven't reviewed yet is a real security risk.
Now, new folders open in Restricted Mode by default. You get a simple banner at the top letting you know you're in safe-browsing mode. You can explore the files, read the code, and when you're ready and confident, you click to trust the folder.
No forced interruption. No accidentally trusting something sketchy before you've looked at it.
The security.workspace.trust.startupPrompt setting defaults to never now instead of once. If you want the old prompt-on-first-open behavior back, set it to once in your settings.
Additionally, the Trust Parent button was removed from the Workspace Trust editor. It looked nearly identical to the regular Trust button but trusted the entire parent folder instead of just the current one ā which was easy to click by mistake. You can still trust a parent folder manually by adding it to the Trusted Folders list.
š° New VS Code Blog Landing Page
Small but worth mentioning: the VS Code website now has a proper blog landing page instead of dropping you directly on the latest post. Previous posts were easy to miss before this. Now you can see recent posts at a glance and browse the full archive.
The documentation table of contents also got restructured ā all agent docs are grouped under "Agents," code editing and config docs are under "Editor," and language/extension docs now live in their own dedicated sections instead of being scattered individually.
Quick Comparison: Before vs After
| Feature | Before 1.126 | After 1.126 |
|---|---|---|
| AI cost visibility | Per-turn only | Full session total |
| Model configuration | Two separate dropdowns | Single unified picker |
| Agent chats per session | One at a time | Multiple side by side |
| Opening new folders | Trust dialog immediately | Restricted Mode first |
| Trust Parent button | Present (and easy to misclick) | Removed |
| Code review comments | Local to client session | Stored on agent host |
Best Tips for Getting the Most Out of VS Code 1.126
ā Do check the session cost popover regularly. If you're on a credit plan, keeping an eye on which sessions consume the most tokens helps you stay within budget.
ā Do use multiple chats for parallel workflows. Keep one chat focused on implementation and another on testing or documentation. Rename the tabs so you know what's what at a glance.
ā Do let Restricted Mode do its job. When you open an unfamiliar repo, take a moment to actually read through the code before clicking Trust. That's the whole point of the feature.
ā
Do try the /code-review skill in the Agents window. If you haven't used agent-driven code review yet, the new comment tools make it significantly more usable.
ā Don't ignore the session cost info. It's easy to dismiss it, but if you're on a team plan or personal credit budget, it's genuinely useful context.
ā Don't set startupPrompt back to once out of habit. The new Restricted Mode default is actually better. Give it a week before reverting.
Common Mistakes to Watch Out For
Thinking multiple chats means unlimited context. Multiple chats in a session share the same working context, which is powerful ā but it doesn't mean the model has infinite memory. Heavy sessions still consume tokens and credits.
Forgetting to rename chat tabs. You open three chats, they all say "Chat 1," "Chat 2," "Chat 3," and twenty minutes later you have no idea which one has the database conversation. Rename them immediately. Double-click the tab, type something useful.
Assuming Restricted Mode means no functionality. You can still read files, search code, and navigate the project in Restricted Mode. Extensions and task runners that need to run code are restricted, which is exactly the point. You're not locked out ā you're just safe-browsing.
Looking for the Trust Parent button and panicking. It's gone on purpose. Add a parent folder path manually to the Trusted Folders list if you genuinely need to trust it. That extra step is intentional so you don't do it by accident.
Wrapping Up š
VS Code 1.126 isn't the flashiest release of the year, but it's a thoughtful one. The team took three real pain points ā AI cost opacity, fragmented agent workflows, and the trust-before-you-look security UX ā and made practical improvements to all three.
Session cost tracking, unified model controls, multi-chat sessions, and safer restricted-mode browsing are all features you'll appreciate quietly in your daily work. The kind of release that doesn't make headlines but makes your day smoother.
Update your VS Code now (or go to Help ā Check for Updates if it hasn't rolled out to you yet), and try opening your next unfamiliar repo in Restricted Mode like the responsible developer you are. š
Want more developer content like this? Head over to hamidrazadev.com for more posts, tools, and updates from the dev world. If this post saved you some time or cleared something up, share it with a fellow dev ā it genuinely helps. š
Top comments (0)