What's new in Claude

Added CLAUDE_CODE_DISABLE_MOUSE_CLICKS env var to disable mouse interactions in fullscreen mode while preserving wheel scroll, plus fixes for hook matchers with hyphenated identifiers, voice dictation, and background job management.

Added autoMode.classifyAllShell setting to route all shell commands through the auto-mode classifier, improved MCP authentication handling, and fixed multiple issues with background agents, file operations, and session management.

claude mcp login: authenticate a configured MCP server directly from the shell instead of navigating the interactive /mcp menu — and clear stored credentials later with claude mcp logout.

Shell mode command awareness: running ! npm test now gets an automatic explanation without needing a second prompt.

/rewind after /clear: the /rewind command can now resume a conversation from before /clear was run, recovering context that would previously have been lost.

Background subagent permission prompts: background subagents now surface permission prompts in the main session instead of silently auto-denying, keeping you in control without breaking the flow.

Rate limits raised across the Claude API. Sonnet and Haiku limits now match Opus at every usage tier. Tiers are consolidated from many levels into three — Start, Build, and Scale — with no org receiving lower limits than before. No action required; check your tier in the Claude Console.

Fast mode deprecated for Claude Opus 4.7, with removal on July 24, 2026. After that date, requests with speed: "fast" to claude-opus-4-7 will error. Migrate to fast mode on Claude Opus 4.8.

Anthropic raised Claude Code weekly limits by 50% through July 13, 2026 — live now for Pro, Max, Team, and seat-based Enterprise users. The Week 26 digest is live covering claude mcp login, shell mode command awareness, and /rewind after /clear.

Major June headlines: Claude Fable 5 launched June 9 with state-of-the-art code and vision capabilities; Claude Tag for Slack entered beta, letting teams delegate tasks and connect tools directly via @Claude in channels; Seoul office opened June 17.

Boris shared a thread on his personal Claude Code setup — "surprisingly vanilla," no heavy customization — and noted that in 2026, 100% of his contributions to Claude Code have been written by Claude Code itself.

Thariq posted about how he and the Anthropic team use Claude to stay in the loop and fully understand work being delegated to agents — exploring patterns for keeping humans meaningfully in the loop at scale.