Published: 2026-06-03

Best Claude Code Features Ranked: Top 12 After 500+ Hours of Use

After 500+ hours inside Claude's ecosystem — chat, Cowork, and Claude Code — Nate Herk ranks every feature from D tier to S tier based on how much each one actually changes his day-to-day knowledge work and automation workflows. Rankings are personal and workflow-dependent, but the video surfaces features many users have never tried or are using incorrectly.

Source video

"I Tested Every Claude Code Feature, These 12 Are the Best" by Nate HerkWatch on YouTube →

Feature Tier List

D Tier — Essential but unremarkable

  • CLAUDE.md — Critical to set up, but just groundwork. Run without it and everything feels off.
  • /clear, /compact, resume — Context lifecycle management. Necessary but not exciting.
  • Web search / web fetch — Native browsing works; spawns research tasks immediately.
  • Custom themes (/theme) — Change between auto, dark, colorblind-friendly, or fully custom.
  • File uploads, fast mode, permissions, IDE extension — All real and useful, but nobody's favorite.

C Tier — Useful, just not central to this workflow

  • Voice mode (/voice) — Dictation mode; use Glido for faster voice input instead.
  • Claude Cowork — Good product, but Claude Code does everything Cowork does plus more. Worth starting with if you're non-technical.
  • Interactive connectors — Easy OAuth-style tool connections in chat or Cowork; less necessary once you're in the terminal.
  • Local file access — A major benefit of terminal-based agents; noted as underused by newer users.

B Tier — High value, still new

  • Dynamic workflows — Processes tasks in parallel; token-heavy but very effective. New enough that the best use patterns are still evolving.
  • /deepresearch — Triggers a dynamic workflow automatically for research tasks. Pairs well with dynamic workflows.
  • Git work trees — Run isolated feature branches that never collide; merge back when done. Essential for multi-feature development.
  • /ultrareview — Spins up a cloud-based code review. Wished it had the visual rainbow indicator like other power features.
  • Recap feature — Shows a summary of what a session was working on when you've been idle or have multiple tabs open.

Top Picks (A Tier and above)

  • Google Workspace CLI — Lets Claude Code navigate Sheets, Docs, Drive, Gmail, and Calendar via CLI commands; skills can be built around it.
  • Dispatch — Continue conversations with your Claude agent from your phone; useful when running Cowork routines away from the desk.
  • Claude desktop app — Better project/session management UI than the terminal; easier copy-paste without terminal line-wrapping issues.
  • Claude Design — Shared design system canvas; useful for teams reviewing landing pages and design systems together.
  • Auto mode permissions — Run in auto-approve mode instead of confirming every action.
  • Hooks and effort levels — Advanced automation and task-depth control.
  • Agent view — Dashboard to monitor all running Claude Code terminals at a glance.
  • Prompt caching — Runs transparently in the background; significant token savings you don't even notice.

Key Takeaways

  • CLAUDE.md is D tier not because it's unimportant, but because it's the bare minimum — you need it before anything else works well.
  • Dynamic workflows and /deepresearch are the breakout newer features; high token cost but strong parallel processing output.
  • Git work trees are underused — critical for anyone working on multi-feature codebases or with a team.
  • The Google Workspace CLI is ranked in the top picks specifically for knowledge work and automation, not software engineering.
  • /ultrareview and /deepresearch are slash commands — type them directly in the Claude Code terminal to invoke.

Commands & Code Mentioned

/voice          # Enable voice dictation mode
/clear          # Clear conversation context
/compact        # Compact/summarize context
/theme          # Change Claude Code theme (dark, auto, colorblind, custom)
/deepresearch   # Trigger dynamic research workflow automatically
/ultrareview    # Launch cloud-based code review
/context        # Ask questions without interrupting the active agent task

Related OpenClaw Guides