Stages
All 14 stages, grouped into 4 phases. Each stage has a goal, core topics, something to build, and a checkpoint.
Where to start
Pick the first stage where you cannot pass the checkpoint. Start there and move forward one stage at a time.
Phase 1 — Foundations
- Stage 0 - Orientation — what agents are and how to use this site
- Stage 1 - Prerequisites — backend, APIs, Git, terminal, JSON, streaming
- Stage 2 - LLM Fundamentals — tokens, context, models, cost, controls
- Stage 3 - Prompt Engineering — instructions, examples, prompt testing
Phase 2 — Agent Core
- Stage 4 - Agent Fundamentals — agent loop, ReAct, planning, stop rules
- Stage 5 - Tools and Actions — tool design, schemas, function calling
- Stage 6 - MCP — connect tools with the Model Context Protocol
- Stage 7 - RAG and Memory — retrieval, embeddings, memory
- Stage 8 - Agent Architectures — ReAct, RAG, planner-executor, routing
- Stage 9 - Building Agents — direct APIs first, then frameworks
Phase 3 — Systems
- Stage 10 - Multi-Agent Systems — supervisor-worker, handoffs, A2A
- Stage 11 - Evaluation and Observability — tests, metrics, tracing
- Stage 12 - Security and Ethics — prompt injection, permissions, red teaming
Phase 4 — Production
- Stage 13 - Production Deployment — APIs, Docker, CI/CD, monitoring, cost