Claude Code's Experimental Swarm Command with Opus 4.6

A quick look at the Claude Code changelog for version 2.1.32, which introduced several notable features including the experimental Agent Teams (swarm) capability powered by Opus 4.6.

What Shipped in 2.1.32

The changelog highlights:

• Claude Opus 4.6 is now available as a model option.
• Agent Teams added as a research preview feature for multi-agent orchestration, enabled via CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1.
• Automatic memory. Claude now automatically records and recalls memories across sessions.
• "Summarize from here" added to the message selector for managing long conversations.
• Skills directory. Skills defined in .claude/skills/ are now recognized within additional contexts.
• Various fixes for file completion, bash tool error handling, and context window budget scaling.

Key Takeaways

• Claude Code 2.1.32 brought Opus 4.6, agent teams, and automatic memory in a single release
• The swarm/agent teams feature enables multi-agent orchestration through a single environment variable flag
• Automatic memory means Claude retains context across sessions without manual configuration