Descriptions:
Bart Slodyczka delivers a structured 22-minute walkthrough of Claude Code’s agent teams feature, introduced alongside the Opus 4.6 model. The video begins by mapping out the three distinct operating modes in Claude Code — default, sub-agent, and agent teams — explaining how each handles session isolation and token consumption differently. Agent teams assign each team member its own dedicated Claude Code session, enabling genuinely parallel, independent work that sub-agents cannot achieve within a single shared session.
The setup section covers everything needed to get started: updating to the latest Claude Code version, editing settings.json to enable the `claude_code_experimental_agents` flag and tmux split panes, and launching a tmux session before starting Claude to ensure each agent gets its own terminal window. Bart introduces TeamUx as a management layer on top of this, and walks through the full session lifecycle — spinning up a team, interacting with individual members via keyboard shortcuts, and cleanly shutting everything down.
The video also breaks down Opus 4.6’s two context window tiers (200k and 1M tokens), the associated pricing shift to $10/$3,750 per million input/output tokens above 200k, and the effort toggle (low, medium, high) that controls reasoning depth and token spend. A closing segment on Claude Code skills files shows how to codify team configurations for reuse across future projects.
📺 Source: Bart Slodyczka · Published February 13, 2026
🏷️ Format: Tutorial Demo
