Claude Code

CLI Agent, CLAUDE.md & MCP Integration

Claude Code

An Agentic CLI for Software Engineering

Claude Code is a command-line tool that acts as an AI agent in your terminal. It can read your codebase, write files, run commands, manage git, and execute multi-step tasks autonomously — all while respecting your project's conventions through CLAUDE.md.

Core Claude Code capabilities
Capability	How It Works
File operations	Read, create, edit, and search across your entire project
Terminal commands	Run builds, tests, installs — sees output and reacts
Git integration	Stage, commit, create branches, open PRs
MCP servers	Connect to external tools, docs, APIs
Plan mode	Outline approach before writing code

# CLAUDE.md — Project Instructions for Claude Code

## Commands
pnpm dev    # Run dev server
pnpm build  # Build for production

## Architecture
- Next.js 16 App Router with Server Components
- Tailwind CSS v4 (no tailwind.config)
- TypeScript strict mode

## Conventions
- Server Components by default
- 'use client' only when needed
- Data fetching in Server Components
- Server Actions for mutations
- All images use next/image

Setup: Install with npm install -g @anthropic-ai/claude-code, then run claude in your project directory. Create a CLAUDE.md at the repo root for project-specific instructions.

Agent Workflow & Plan Mode

Autonomous Execution with Human Oversight

Claude Code operates as a full agent — give it a task and it autonomously plans, executes steps, reads results, and iterates. Use /plan to make it outline the approach first.

$ claude
> Add a /dashboard page that shows task statistics
  grouped by status and priority. Include a chart.

# Claude autonomously:
# 1. Reads src/api.ts to understand the data model
# 2. Reads existing pages for layout patterns
# 3. Creates src/app/dashboard/page.tsx
# 4. Creates src/components/stats-chart.tsx
# 5. Adds dynamic import for chart (code splitting)
# 6. Adds metadata export
# 7. Updates layout.tsx navigation
# 8. Runs pnpm build — catches a type error
# 9. Fixes the type error
# 10. Runs pnpm build again — success
# 11. Shows you a summary of all changes

Tip: Use /plan for any task touching 3+ files. It saves time by catching wrong approaches before code is written.

Skills & MCP Integration

Extending Claude Code's Capabilities

Skills are reusable slash commands that expand into detailed prompts. MCP servers connect Claude Code to external tools and data sources.

/commit — Analyzes changes, writes a conventional commit message
/review — Reviews code for bugs, security issues, and best practices
Custom skills — Create your own in .claude/skills/ for team-specific workflows

// In .claude/settings.json or project config:
{
  "mcpServers": {
    "nextjs-docs": {
      "command": "npx",
      "args": ["-y", "@anthropic-ai/mcp-server-fetch"]
    },
    "postgres": {
      "command": "npx",
      "args": ["-y", "@anthropic-ai/mcp-server-postgres"],
      "env": {
        "DATABASE_URL": "postgresql://..."
      }
    }
  }
}

Key Insight:Claude Code's strength is end-to-end task execution — it doesn't just generate code, it creates files, runs builds, fixes errors, and verifies the result. The combination of CLAUDE.md + skills + MCP makes it deeply aware of your specific project.