Claude Code Worktrees in 7 Minutes

Description

Anthropic has released Git Wortrees in Claude Code, bringing a feature previously available in the Claude Desktop app directly into the CLI. The script explains Git worktrees as a way to check out multiple branches of the same repository simultaneously in different directories while sharing the same Git data, and shows the prerequisites to use them (a Git-initialized directory with at least one commit). It demonstrates running `claude-worktree` in multiple terminals to create isolated working directories and make different HTML changes in parallel, then examines the generated folder structure with unique worktree paths. The episode also shows how to use worktree isolation with Claude Code subagents to parallelize tasks—spawning five subagents to create five different SaaS landing page variations—and highlights benefits like avoiding conflicts and offloading context from the main thread. It covers creating and configuring subagents via natural language and via files with front matter (name, description, model such as Haiku, and the new `isolation: worktree` setting), where subagent definitions can live globally or within a project under `.claude/agents`, and notes that tool access can be whitelisted (including MCP tools). The video closes by suggesting use cases such as testing, migrations, and exploring divergent implementation paths, and thanks the Anthropic team before a standard like/comment/subscribe outro. 00:00 GI Worktrees land in Claude Code (what’s new) 00:26 What Git worktrees are & why agents benefit 00:56 Hands-on setup: initialize repo, commit, and run worktree in CLI 01:24 Demo: two parallel edits in separate worktrees (no conflicts) 01:42 Where worktrees live: .cloud folder, paths, and git data sharing 02:08 Scaling with subagents: spawn 5 landing-page variants in parallel 03:48 More use cases: testing, migrations, and exploring divergent ideas fast 05:18 Create a subagent via natural language + enable worktree isolation 05:40 Subagent file anatomy: front matter, isolation flag, and tool whitelists 07:01 Wrap-up and closing thoughts