By VS Code 1.126 shipped on June 24 with session-level cost tracking for Copilot chat — Microsoft’s most direct acknowledgment yet that Copilot’s June 1 token billing switch blindsided a lot of developers. If you have been watching your AI Credits drain faster than expected, the answer is now one click away in the editor.
First, the billing context
GitHub Copilot switched from flat-rate request billing to token-based AI Credits on June 1. The math hit hard: one developer reported a projected $180 bill on their first day. Heavy agentic users saw costs jump from $29 to $750 per month, and in extreme cases from $50 to $3,000. The official GitHub Community thread drew 400+ comments and nearly 900 downvotes. TechCrunch summed up the sentiment with a headline that pulled no punches.
The credit economics are simple but unforgiving. One AI Credit equals $0.01. A Pro+ subscription ($70/month) includes 7,000 credits — enough for roughly five to seven days of heavy agentic work. A typical mixed day runs $10–$14 in credits. A single complex Opus session can cost $0.50–$2.00. A developer doing a full day of agentic work burned through 360 credits in one sitting. None of this was obvious before VS Code started surfacing the numbers. Microsoft first responded with a spend meter in VS Code 1.125 on June 22. With 1.126, they go further.
Session-level cost tracking
The main addition in VS Code 1.126 is visibility into the total cost of a chat session, not just individual turns. That distinction matters more than it sounds. Before, you could see that each message cost some credits, but there was no easy way to know that the three-hour debugging session you just closed burned 400 credits — a meaningful chunk of your monthly budget.
Now you can. The session cost view makes it straightforward to identify which workflows are expensive: long agent runs with premium models, large context windows, repeated tool-call cycles. Once you see the pattern, you can do something about it — switch to a lighter model for exploratory work, break large tasks into smaller sessions, or make a more informed decision about when to reach for Opus versus a cheaper alternative.
Multiple chats per agent session
Separate from the billing work, VS Code 1.126 brings a genuine workflow improvement: the Agents window now supports multiple simultaneous chats within a single session. The chats share working context but maintain independent conversation threads. Start one agent working on a feature while another investigates a bug — simultaneously, in the same session. Conversations now persist across window reloads, which they previously did not. Rename tabs independently by double-clicking, regardless of the session title.
The persistence fix is the quietly important one. Losing agent conversations on window reload was a friction point that discouraged leaving long-running sessions open. That excuse is gone now. Microsoft is treating multi-chat agent workflows as a permanent, stable part of the editor rather than an experiment — which is the right read on where AI-assisted development is heading.
Two smaller changes worth knowing
The model customization controls are now a single picker. Context size and reasoning effort were two separate dropdowns; now they’re one. Minor, but the direction is right — less friction when tuning model behavior.
More significantly: new folders now open in Restricted Mode by default. The security.workspace.trust.startupPrompt setting now defaults to never rather than once. Instead of an immediate trust prompt, you get a trust banner, which is the right default when reviewing code from unfamiliar repositories. The “Trust Parent” button has also been removed — it was causing accidental broad directory trust grants, which is a real security risk in any environment where you’re opening third-party code.
The bottom line
VS Code 1.126 is a meaningful quality-of-life release, not a cost fix. The billing model is still the billing model. What it gives you is the visibility you should have had from day one — and a multi-chat agent workflow that is genuinely useful independent of the billing drama surrounding it.
If you are running agentic workflows in VS Code and have not checked your session costs yet, update to 1.126 and spend ten minutes reviewing what you are actually spending. The numbers may change which models you reach for. Code completions and Next Edit Suggestions remain free and unlimited on all plans, which is worth keeping in mind when you’re deciding how to structure your workflow going forward.
