This hub sequences the concepts an analytics engineer encounters in the first hour with Claude Code: installation, authentication, project exploration, and first useful output. Claude Code reads dbt_project.yml, scans model directories, and infers conventions from existing SQL rather than matching on filenames.
The Setup Path
1. Installation and CLI Basics
Claude Code CLI Basics covers installation, launching Claude Code, and the essential slash commands (/help, /clear, /compact, /model). Start here if you’ve never run claude before. The permission model — where Claude asks before reading, writing, or executing — is the foundation of the trust relationship.
2. Authentication
Claude Code Authentication Options explains the two paths: OAuth with a Claude Pro/Max subscription (simpler, start here) and API keys (for automation and CI/CD later). The key detail most people miss: if ANTHROPIC_API_KEY is set in your environment, it overrides your subscription silently.
3. Model Selection
Claude Code Model Selection for Analytics Work covers when to use Sonnet vs Opus. Default to Sonnet for daily work — base models, tests, documentation, simple debugging. Escalate to Opus for complex incremental model debugging, nested macro refactoring, and multi-file architectural decisions.
4. Shell Commands Inside Claude Code
Claude Code Bang Prefix for Shell Commands introduces the ! prefix for running commands directly without consuming tokens. Essential for GCP authentication (!gcloud auth login), quick checks (!dbt ls -s tag:daily), and git operations (!git checkout -b feature/new-model).
5. Status Line
Claude Code Status Line Configuration shows how to display your git branch, active model, and context usage at a glance. A one-minute setup that prevents ambient confusion about which branch, which model, and how much context you’ve consumed.
Your First Useful Output
6. Base Model Generation
Base Model Generation with Claude Code is the natural first task. Point Claude at a source table and ask it to create a base model following your existing patterns. The output follows your conventions because Claude reads your code and replicates the structure — not because you described the pattern in the prompt.
Create a base model for [your_source].[your_table] following the patterns in my existing base modelsReview the diff. Check naming conventions. Verify the patterns match. If something’s off, tell Claude and it adjusts.
What to Skip (For Now)
7. CLAUDE.md
CLAUDE.md as Project Memory is the single highest-impact configuration for Claude Code — but you don’t need it in your first hour. Claude already reads your project files and discovers patterns on its own. Create a CLAUDE.md after a few sessions, when you notice Claude making the same mistake twice. Grow it from real friction, not from guesses.
What Comes Next
After the first hour:
- Write tests — Claude identifies primary keys and generates uniqueness, not-null, and accepted_values tests.
- Debug an error — paste an error message and Claude traces through the model and its dependencies.
- Document a model — Claude reads the SQL and writes column descriptions that explain what fields mean.
The Claude Code for dbt Development hub covers the full set of workflows: TDD, documentation, refactoring, slash commands, and hooks.
The Claude Code ROI for Analytics Engineers note gives the realistic timeline: several weeks of regular use before the tool consistently saves time.