By VS Code 1.123 shipped June 3 with three features that change how long-running agent work actually holds together. Your chat sessions now follow you across machines. Context windows hit 1 million tokens. And a new research agent can audit your entire codebase without touching a file. Each one solves a real frustration. All three land in the same release.
Session Sync: Your Agent Work Follows You
The most impactful addition is session sync. Previously, switching from your laptop to your desktop mid-task meant starting over — the agent session was gone. Now VS Code syncs your chat sessions automatically to your GitHub account. Each session captures the full conversation, every file you edited, repo context (branch, timestamps), and any PRs or issues referenced along the way.
It’s on by default. You get it without touching a setting. If you want to keep sessions local only, set chat.sessionSync.enabled to false. Enterprise teams can manage it via the “Store local sessions in the Cloud” GitHub.com policy. Synced sessions are private — no one else can see them unless you share. See the session sync documentation for the full configuration options.
Alongside sync comes /chronicle, a set of slash commands for querying that history:
/chronicle:standup— generates a standup report from the last 24 hours of coding, grouped by feature and branch with file lists and PR links/chronicle:tips— analyzes a week of usage and surfaces personalized workflow improvements/chronicle [query]— free-form natural language search across session history; try the /chronicle docs for query examples
The standup generator is the kind of thing you’ll use immediately. For developers juggling multiple repos daily, the “what did I actually do yesterday?” mental tax is real. /chronicle:standup kills it.
1M Token Context Windows: No More Mid-Session Truncation
Context window truncation mid-agent session was the top complaint in VS Code GitHub issues. 1.123 addresses it directly: VS Code now supports 1 million token context windows for compatible models, including Claude Opus 4.7 and GPT-5.5.
One million tokens is approximately 750,000 words — enough to hold an entire large codebase, run an agent for hours, and stop manually managing what fits. No special configuration is required. Compatible models get the expanded limit automatically in 1.123. Per-token pricing stays flat, so a 900K-token request costs the same rate per token as a 9K one. The only caveat: larger contexts do increase AI credits consumption under usage-based billing plans, so monitor your usage if you’re on a metered plan.
Research Agent: Investigate Without Touching Code
The research agent is exactly what it sounds like: an agent that investigates and reports, and does nothing else. It is read-only by design. Run /research [your question] in a Copilot CLI (local) session and it produces a well-cited Markdown report drawing from your codebase, relevant GitHub repositories, and the web.
The read-only constraint is the right call. For onboarding onto an unfamiliar codebase or running an architectural audit before a major refactor, you want depth without risk. The research agent delivers the first without the second.
One caveat: it’s currently in preview and only available in Copilot CLI (local) sessions in Insiders. If you’re on VS Code Stable, you’ll need to wait for 1.124 to get it. Track the GitHub Copilot May releases changelog for the stable rollout date.
Safer Extension Updates and Browser Improvements
1.123 also adds a 2-hour automatic delay before extension updates install. Trusted publishers still push instantly, but the buffer gives teams time to catch broken releases before they propagate. Given how many developers have been burned by a bad auto-update mid-sprint, this is a practical change that will prevent real headaches.
The integrated browser gets a remodeled address bar with favorites support and two new screenshot commands — “Add Area Screenshot to Chat” and “Add Full Page Screenshot to Chat” — that let you drop browser screenshots directly into agent conversations.
What to Do Now
- Update VS Code. Help > Check for Updates, or download from code.visualstudio.com.
- Try
/chronicle:standupafter your next session to see what it captures. - Switch to a compatible 1M context model (Claude Opus 4.7 or GPT-5.5) if you run long agent sessions. No config changes needed — just select the model.
Session sync is the kind of feature that changes daily habits. The 1M context window removes a real ceiling. Check the full VS Code 1.123 release notes for the complete list of changes.
