March 26, 2026 · 5 min read

# 7 Claude Code Features You're Not Using (But Should Be)

I've been using Claude Code daily for 8 months. Most people use it like a glorified autocomplete. They're leaving 90% of the power on the table.

Here are 7 features that changed how I ship code.

1 CLAUDE.md — Your Project's Brain

Most people don't know Claude Code reads a CLAUDE.md file in your project root. This is where you store:

  • Architecture decisions — "We use React Query, not SWR"
  • Coding conventions — "Always use TypeScript, never any"
  • Context about the domain — "This is a fintech app, security matters"

Create this file once. Claude remembers it forever. Every code suggestion respects your rules.

2 /compact — Context Magic

When your conversation gets long, use /compact. It summarizes everything important and frees up context window. You get a fresh start without losing the thread.

This is crucial for long coding sessions where you're iterating on complex features.

3 Agent Mode — Let It Run

Type # to enter agent mode. Claude can now:

  • Run terminal commands
  • Edit multiple files
  • Run tests and fix failures
  • Install packages

I use this for refactoring. "Extract all the auth logic into a separate module." Then I go make coffee. When I come back, it's done.

4 The Todo List You Didn't Know About

Claude Code maintains an internal todo list during complex tasks. You can reference it:

What's left on the todo list?

This keeps multi-step tasks on track. No more "oh I forgot to update the tests."

5 Image Understanding

Paste screenshots directly into Claude Code. It can:

  • Read error messages from screenshots
  • Understand UI mockups and generate code
  • Debug visual issues from screenshots

I paste UI designs and get working components in seconds.

6 Parallel Agent Spawns

You can spawn multiple Claude agents in parallel. Each works independently on different parts of your codebase.

Spawn an agent to handle the API routes
Spawn another agent for the frontend components

Two tasks, same time. This is how you 10x your output.

7 MCP Servers — Extend Claude's Reach

Model Context Protocol (MCP) servers let Claude Code connect to:

  • PostgreSQL databases — Query and modify data
  • GitHub — Create PRs, review code, manage issues
  • Slack/Discord — Send messages, read channels
  • Browser automation — Navigate, scrape, test

Once configured, Claude becomes a full-stack assistant. Not just coding — doing.

The Pattern

Notice something? These aren't just "write code faster" features. They're "think and act independently" features.

Claude Code isn't a tool. It's a teammate that works 24/7, doesn't get tired, and remembers everything.

The developers who win aren't the ones who type fastest. They're the ones who set up systems that work while they sleep.

Want the Full Cheat Sheet?

I compiled every Claude Code tip, shortcut, and workflow into a 15-page cheat sheet. 47 commands, 12 workflows, 3 custom configs.

What's Next

Start with CLAUDE.md. That alone will change how you work. Then try /compact on your next long session.

The others you'll pick up naturally as you hit walls. "There has to be a better way" — there usually is.

Related: Claude Code Cheat Sheet 2026 · How I Automated My Life with OpenClaw