The Foundation

What is Claude Code?

Claude Code is Anthropic's terminal-based AI coding assistant. It runs in your shell, reads and writes files, executes commands, manages git, runs tests, and debugs code — all through natural language. You describe what you want; Claude Code does the work.

Unlike browser-based AI assistants, Claude Code operates directly on your filesystem with full access to your project. It sees your code, understands your dependencies, and makes changes in place. No copy-pasting, no context windows limited to a single file.

Reads, edits, and creates files directly in your project directory
Runs shell commands, tests, and build scripts
Manages git operations (commits, branches, diffs)
Supports multiple Claude models (Opus, Sonnet, Haiku) for different task complexities
Outputs structured stream-JSON for programmatic integration

# Install Claude Code npm install -g @anthropic-ai/claude-code # Navigate to your project cd ~/my-project # Start Claude Code claude # Now talk naturally: "Read the auth module and add rate limiting to the login endpoint. Use a sliding window of 5 attempts per 15 minutes per IP." # Claude Code will: # 1. Read the existing auth files # 2. Identify the login endpoint # 3. Write the rate limiter # 4. Update tests # 5. Run the test suite

Fundamentals

Claude Code basics every developer should know

Core patterns for productive single-agent development.

1

Read Before Writing

Always ask Claude Code to read existing code before modifying it. "Read src/auth.ts and explain the token validation flow" gives Claude full context. Blind edits waste tokens and produce misaligned code.

2

Be Specific With Tasks

"Fix the bug" is vague. "The login endpoint returns 500 when the email contains a plus sign — fix the validation regex in src/validators.ts" gives Claude Code everything it needs to produce the right fix on the first try.

3

Use CLAUDE.md for Context

Drop a CLAUDE.md file in your project root with conventions, architecture notes, and do-not-touch rules. Claude Code reads it automatically at the start of every session. It's your persistent context that survives between conversations.

4

Choose the Right Model

Use Opus for architecture decisions and complex refactoring. Sonnet for standard feature development and bug fixes. Haiku for quick lookups, formatting, and boilerplate. Match model cost to task complexity.

5

Let It Run Tests

After every change, tell Claude Code to run your test suite. "Run npm test and fix any failures" creates a tight feedback loop. Claude Code reads test output, identifies failures, and fixes them iteratively.

6

Commit Incrementally

Ask Claude Code to commit after each meaningful change. Small, focused commits are easier to review and revert. "Commit with a descriptive message" keeps your git history clean without manual effort.

The Ceiling

Where single-agent Claude Code hits its limit

Claude Code is powerful as a solo agent, but real projects need parallel work. One agent can't build the database schema, API routes, frontend components, and test suite simultaneously. You end up sequencing everything through a single conversation, waiting for each task to complete before starting the next.

The problems compound with project size:

Context windows fill up — Claude loses track of earlier work in long sessions
No isolation — one agent can accidentally overwrite another agent's files
No recovery — if the session crashes, you restart from scratch with no handoff
No visibility — you can't monitor multiple tasks running in parallel
No cost control — no way to set budgets or track spending per task

This is where multi-agent orchestration changes the game.

// Single agent: sequential, slow // One agent does everything: claude "Build the database schema" // ...wait 10 minutes... claude "Now build the API routes" // ...wait 15 minutes... claude "Now build the frontend" // ...wait 20 minutes... // Total: 45 minutes, sequential // Multi-agent with chAIrman: // 4 agents work in parallel batch_hire({ agents: [ { role: "DB Engineer" }, { role: "Backend Lead" }, { role: "Frontend Lead" }, { role: "QA Engineer" } ] }) // Total: ~15 minutes, parallel // With dependency ordering

Scale Up

Multi-agent Claude Code with chAIrman

The step-by-step path from one agent to a full AI workforce.

1

Install chAIrman

Download the binary, add it to your MCP config, and restart Claude Desktop. You now have access to 54 orchestration tools. Claude Desktop becomes your CEO — it plans, hires, and manages agents through natural language.

2

Create a Backlog

Tell Claude Desktop what you want to build. It creates a structured backlog with milestones and tickets. Each ticket has a description, files to touch, files off-limits, and success criteria. You approve the plan before any agent is hired.

3

Hire Specialized Agents

Each ticket gets one agent. Use templates for common roles (frontend-lead, backend-lead, qa-engineer) or write custom job descriptions. Agents receive the project's CLAUDE.md, relevant skills from the 857-skill library, and their specific ticket context.

4

Wire Dependencies

Use depends_on to order agents. The QA agent waits for the backend and frontend agents to finish. chAIrman handles the queuing, cycle detection, and auto-launching when dependencies resolve.

5

Monitor Progress

Open the dashboard at localhost:3456 to watch agents work in real time. See the kanban board, pipeline visualization, cost tracking, and live terminal output. Desktop notifications fire when tasks complete or budgets are nearing limits.

6

Build Institutional Memory

Fire agents when they're done — they're archived with full history. Next time you need a similar role, rehire a veteran with rehire_veteran. They come back with all their previous context, reducing ramp-up time.

54

Orchestration Tools

857

Auto-Injected Skills

10

Agent Templates

3

Model Tiers

Common Questions

Claude Code tutorial FAQ

How do I install Claude Code?

Install Claude Code globally via npm: npm install -g @anthropic-ai/claude-code. You need Node.js 18 or higher and an Anthropic API key set as the ANTHROPIC_API_KEY environment variable. After installation, run claude in any project directory to start.

What's the difference between Claude Desktop and Claude Code?

Claude Desktop is a GUI application for conversational AI — you chat with Claude through a window. Claude Code is a terminal tool that operates directly on your filesystem: reading files, writing code, running commands. With chAIrman, Claude Desktop becomes the "CEO" that orchestrates multiple Claude Code "agents" working on your project simultaneously.

How much does multi-agent development cost?

Costs depend on the models and task complexity. Sonnet tasks typically cost $0.10–$0.50 each. Opus tasks run $0.50–$3.00. chAIrman tracks costs per-agent and per-task, and you can set project budgets to cap spending. A typical 4-agent feature build costs $2–$8 total.

Can I use Claude Code without chAIrman?

Absolutely. Claude Code works great as a standalone tool for single-agent development. chAIrman adds value when you need multiple agents working in parallel, dependency pipelines, cost tracking, auto-retry, skills injection, and the monitoring dashboard. Start with Claude Code alone and add chAIrman when you need to scale.

What languages and frameworks does Claude Code support?

Claude Code is language-agnostic. It works with any programming language, framework, or toolchain that runs in a terminal. TypeScript, Python, Rust, Go, React, Next.js, Django, Rails — Claude Code reads and writes all of them. chAIrman's skills library includes specialized skills for common stacks.

How do agents avoid overwriting each other's work?

chAIrman's ticket system enforces file ownership. Each ticket defines files_to_touch and files_off_limits. The file claims system tracks which agent owns which file in real time. Two agents never edit the same file simultaneously, preventing merge conflicts and wasted work.

Claude Code Tutorial: From Solo to Multi-Agent