Claude Code

Nate Layman

Follow Along

tinyurl.com/claudecodedemo

Agentic vs Non-Agentic AI

Non-Agentic

Traditional AI Chat Interfaces

  • Response generation only
  • No environment access
  • Manual implementation required
  • Text-based interaction

Agentic

Environment-Integrated AI

  • Direct file system access
  • Tool execution capabilities
  • Automated task completion
  • Code modification abilities

What is Claude?

Anthropic’s large language model

  • Advanced reasoning capabilities
  • Long context windows (200K+ tokens)
  • Multiple model tiers
  • Strong coding abilities

Why Claude Code?

Traditional AI Workflow:

Copy code → Paste to ChatGPT → Copy solution → Test → Breaks → Repeat…

Claude Code:

Direct file access, full project context, automated execution

No more copy-paste. No more context loss.

The Power of Multi-File Context

Traditional AI sees snippets. Claude Code sees your entire project:

  • Understanding relationships between files
  • Cross-file dependencies and imports
  • Holistic refactoring across the codebase

Example: Rename a function used in 47 files across 12 modules - Claude Code updates everything consistently.

Claude Code: Productivity Multiplier

3x productivity gains with proper AI assistance

  • Refactor functions across 47 files in 5 minutes
  • Full project context and multi-file awareness
  • Automated test generation and documentation
  • File organization, project maintenance, and housekeeping

Terminal-Based Architecture

  • Works with any editor: VS Code, IntelliJ, Vim, Emacs
  • Lightweight command-line interface
  • No IDE-specific plugins required
  • Preserves existing workflows

What Claude Code Sees

  • Full project context
  • All source files
  • Tests and documentation
  • Configuration files

Security Benefits

  • Project-root bounded access
  • No access to parent directories
  • Focused, relevant assistance
project-root/
├── src/         ← Can see
├── tests/       ← Can see
├── docs/        ← Can see
├── package.json ← Can see
└── ../other/    ← Outside

GitHub Integration Made Simple

Alternative to “delete the project and download a fresh copy”

Git Operations: - Commit message generation - Branch management - Pull request creation - Issue tracking integration

Claude Code + GitHub Projects

Powerful project management synergies:

  • Automate issue creation from code analysis
  • Update project boards based on code changes
  • Generate sprint summaries from PRs and issues
  • Create technical tasks from high-level goals
  • Keep documentation in sync with project status
  • Interrogate commit history to recall what you were working on

Claude Code Pricing

API Access

Pay per token

  • $3-15 per million tokens
  • Variable monthly costs
  • Usage-dependent

Subscription

Fixed monthly fee

  • Pro: $17-20/month
  • Team: $150/person/month
  • Predictable costs

Installation Process

# Install via npm
npm install -g @anthropic/claude-code

# Authenticate
claude-code auth login

# Initialize in project directory
cd your-project
claude-code

Standard command-line installation and setup.

Configuration & Customization

CLAUDE.md Files

Auto-loaded project instructions at repo root

  • Bash commands and testing instructions
  • Code style guidelines and repository etiquette
  • Developer environment setup

Terminal Settings

  • /terminal-setup - Configure Shift+Enter for multiline
  • /permissions - Manage tool permissions
  • /output-style - Change interaction style

Interaction Modes

Press Shift+Tab to cycle through three modes:

Default Mode Claude suggests changes and waits for your permission

Auto-Accept Mode Claude executes changes autonomously without waiting

Plan Mode Claude creates a comprehensive plan without making any changes - Read-only research and analysis phase - Review the plan before execution - Accept with auto-edits or manual edits

Press Escape to interrupt Claude at any time

Best Practices

Recommended Workflow:

  1. Define the end state - Specify what success looks like
  2. Create verification - Design tests that prove success
  3. Find the path - Iteratively code until tests pass

Optimization Tips

  • Be specific in instructions
  • Course correct early
  • Use /clear to manage context

Gordon’s Program

From Douglas Adams’ “Dirk Gently’s Holistic Detective Agency”

“Gordon’s great insight was to design a program which allowed you to specify in advance what decision you wished it to reach, and only then to give it all the facts. The program’s task, which it was able to accomplish with consummate ease, was simply to construct a plausible series of logical-sounding steps to connect the premises with the conclusion.”

Traditional Coding

  1. Write code
  2. Run manually
  3. Find bugs
  4. Fix issues
  5. Repeat

Test-Driven Development

  1. Define expected outcome
  2. Write tests first
  3. Iterate until tests pass

Why Working Backwards Works

Similar to simulating data for power analysis:

  1. Start with your expected model
  2. Generate data (or code) to fit the model
  3. Refine the approach

Claude Code can do 2 and 3 for you!

Live Demo: File Organization

What we’ll demonstrate:

  1. Create new branch and switch to it
  2. Create GitHub issue: “organize notes”
  3. Use Claude Code to organize messy files
  4. Commit changes with Claude Code
  5. Push branch and open PR with Claude Code
  6. Merge PR with Claude Code