Member-only story
You upgraded Claude Code at the end of June and got a version bump in your terminal. You read the first three lines of the changelog, noticed it mentioned “background agent fixes,” and moved on. That’s understandable — the changelog for v2.1.198 through v2.1.206 spans hundreds of bullet points, and the signal is genuinely buried in the noise.
What you missed is a cluster of features that substantially change the day-to-day shape of working with Claude Code at the team level. Not polish. Structural improvements to how you debug broken configs, how MCP authentication works at enterprise scale, how you share session output with reviewers, and how background agents behave when nobody is watching.
This article surfaces the ten highest-impact additions from the June–July 2026 release window: what each one does, how to use it, and where the trade-offs sit.
Source note:All feature claims are drawn from the official Claude Code changelog atgithub.com/anthropics/claude-code, theClaude Code docs changelog, the Anthropic blog, and the MCP blog at blog.modelcontextprotocol.io. Version numbers are noted where relevant.
A note on terminology #
Three related extension concepts appear throughout this article: