Overview

Claude Code introduces multi-agent collaboration capabilities alongside memory persistence and improved session management. The latest releases focus on enabling teams of AI agents to work together on complex tasks while maintaining context across conversations.

Changes by Version

v2.1.33

  • Fixed agent teammate sessions in tmux to send and receive messages
  • Fixed warnings about agent teams not being available on your current plan
  • Added TeammateIdle and TaskCompleted hook events for multi-agent workflows
  • Added support for restricting which sub-agents can be spawned via Task(agent_type) syntax
  • Added memory frontmatter field support for agents with user, project, or local scope
  • Added plugin name to skill descriptions and /skills menu for better discoverability
  • Fixed interruption of thinking phase when submitting new messages during extended thinking
  • Fixed API error during mid-stream aborts with whitespace and thinking blocks
  • Fixed API proxy compatibility where 404 errors no longer trigger non-streaming fallback
  • Fixed proxy settings not applying to WebFetch and HTTP requests on Node.js build
  • Fixed /resume session picker showing raw XML markup instead of clean titles
  • Improved error messages for API connection failures with specific causes
  • Added error surfacing for invalid managed settings
  • Added VSCode remote sessions support for OAuth users to browse claude.ai sessions
  • Added git branch and message count to VSCode session picker with branch name search
  • Fixed VSCode scroll-to-bottom under-scrolling on session load and switch

v2.1.32

  • Added Claude Opus 4.6 availability
  • Added research preview agent teams feature for multi-agent collaboration (requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1)
  • Added automatic memory recording and recall as Claude works
  • Added “Summarize from here” to message selector for partial conversation summarization
  • Added automatic skills loading from .claude/skills/ in additional directories
  • Fixed @ file completion showing incorrect relative paths from subdirectories
  • Updated –resume to re-use –agent value from previous conversations by default
  • Fixed Bash tool “Bad substitution” errors with JavaScript template literals in heredocs
  • Added skill character budget scaling with context window (2% of context)
  • Fixed Thai/Lao spacing vowels rendering in input field
  • Fixed VSCode slash commands executing incorrectly when pressing Enter with preceding text
  • Added VSCode spinner when loading past conversations list