Skip to main content
Official reference — The official parallel execution docs cover git worktrees, the /batch command, and worktree isolation mechanics. This page adds the opinionated workflow.

​
Why Parallel?

One agent handles one task at a time. While it’s writing tests, it’s not implementing the next feature. The fix: run multiple agents. This is the single biggest productivity unlock in agentic engineering.

​
The Cascade

The workflow:
  1. Open a new terminal tab for each task
  2. Arrange tabs left-to-right in priority order
  3. Kick off each agent with a clear, scoped prompt
  4. Sweep left to right to review output as agents finish
Cap at 3-4 active sessions — Beyond that, your context-switching cost as the reviewer starts outweighing the throughput gain. You become the bottleneck, not the agents.

​
Shell Aliases for Rapid Switching

Set up short aliases to jump between worktrees and launch sessions fast: 
# Quick directory jumps
alias za="cd ~/projects/my-app"
alias zb="cd ~/projects/my-app-feat-a"
alias zc="cd ~/projects/my-app-feat-b"

# Quick Claude launches
alias c="claude"
alias ch="claude --with-chrome"  # Claude with browser MCP
Small friction reductions compound when you’re switching between 3-4 sessions throughout the day.

​
How Many Is Too Many?

SessionsVerdict
1Baseline — you’re serializing work
2-3Sweet spot for most engineers
4-5Effective if tasks are independent and well-scoped
6+Diminishing returns — you can’t review fast enough
The constraint isn’t compute. It’s your ability to review and integrate the output.

​
Combining with Session Management

Each worktree should be self-contained:
  • Its own Claude session
  • Its own HANDOFF.md if you need to pause and resume
  • Its own branch with regular commits
When an agent finishes in one worktree, review the output, merge the branch, and clean up the worktree. Then spin up a new one for the next task.

​
Going Deeper

For more sophisticated multi-agent coordination, see:
← Prev: Session Management · Next: Prompting Craft →