Overview

Claude Code v2.1.63 delivers significant improvements to memory management and developer workflow efficiency. The release fixes numerous memory leaks that plagued long-running sessions while adding new slash commands and workflow enhancements that streamline code interaction.

View Original

Changes by Version

v2.1.63

  • Added /simplify and /batch bundled slash commands for streamlined code operations
  • Added HTTP hooks for JSON-based integrations instead of shell commands
  • Added “Always copy full response” option to skip code block picker in /copy commands
  • Added project configs and auto memory sharing across git worktrees of the same repository
  • Added manual URL paste fallback for MCP OAuth authentication when localhost redirect fails
  • Added ENABLE_CLAUDEAI_MCP_SERVERS=false env var to opt out from claude.ai MCP servers
  • Added session rename and remove actions to VSCode sessions list
  • Fixed local slash command output appearing as user messages instead of system messages
  • Fixed memory leak in long-running sessions - teammates now properly compact conversation history
  • Fixed memory leak in MCP server fetch caches that grew with frequent reconnections
  • Fixed memory leak in git root detection cache that caused unbounded growth
  • Fixed memory leak in JSON parsing cache during long sessions
  • Fixed memory leak in bash command prefix cache
  • Fixed memory leak when navigating hooks configuration menu
  • Fixed listener leaks in bridge polling loop and MCP OAuth flow
  • Fixed listener leaks in interactive permission handler during auto-approvals
  • Fixed WebSocket listener leak on transport reconnect
  • Fixed file count cache ignoring glob ignore patterns
  • Fixed IDE host IP detection cache incorrectly sharing results across ports
  • Fixed MCP tool/resource cache leak on server reconnect
  • Fixed race condition in REPL bridge causing message ordering issues during initial connection
  • Fixed remote sessions not appearing in VSCode conversation history
  • Fixed /clear not resetting cached skills causing stale content to persist
  • Improved /model command to show currently active model in slash command menu
  • Improved memory usage with subagents by stripping heavy progress payloads during compaction