4 min · Lesson 1 of 8
API keys, prompt testing, usage dashboard
Where you ship anything that uses Claude programmatically.
Tracing, evals, debugging for LLM apps
When you start running agents in production, you'll need traces.
Vector DB for RAG and memory
The default vector DB if you outgrow SQLite for embeddings.
Eval and experiment tracking
When you start A/B testing prompts at scale.