How to Use Subagents in Claude Code

Description

Subagents let you spawn multiple Claude instances—each with their own context window and custom instructions—so you can parallelize work instead of grinding through tasks one at a time. I walk through creating a markdown linting agent from scratch, spawning three parallel instances across 86+ files, and managing foreground/background workflows. Real debugging, real hiccups, real solutions. === TIMESTAMPS === 00:00 - What are subagents? 00:35 - Why context isolation matters 01:00 - Prerequisites and terminal setup 01:30 - /init and CLAUDE.md project memory 02:15 - The /agents interface 03:00 - Project vs. user-level agents 03:45 - Using Claude to write agent instructions 05:00 - Generating the agent 05:30 - Selecting tools and models 07:00 - Understanding agent markdown files 07:45 - Invoking agents via delegation 08:20 - Spawning parallel agents 09:05 - Foreground/background with Ctrl+B 10:00 - Confirmations and autonomy tradeoffs 11:05 - Self-correcting agent behavior 12:15 - Working in parallel 14:00 - Verifying results 15:20 - /context for token visualization 17:00 - Wrap-up === KEY RESOURCES === Claude Code Docs: https://docs.anthropic.com/en/docs/claude-code/overview Subagents Docs: https://docs.anthropic.com/en/docs/claude-code/sub-agents Best Practices: https://www.anthropic.com/engineering/claude-code-best-practices Claude Pricing: https://claude.com/pricing === KEY TAKEAWAYS === → Subagents get isolated context—no pollution of your main chat → Store agents at project or user level for reuse → Ctrl+B backgrounds agents so you keep working → /context visualizes token usage breakdown === ABOUT ME === Tim Warner | Microsoft MVP | 28+ years in Microsoft tech Principal Staff Author at Pluralsight | O'Reilly Live Learning Instructor https://TechTrainerTim.com #ClaudeCode #Anthropic #Subagents #AgenticAI #AIcoding #DevTools #VSCode #ContextEngineering #ClaudeMax #AIworkflow