Agentic AI — One-Stop Guide¶
Everything you need to design, build, and ship autonomous AI systems.
Dedicated Agent Engineering track
For a structured curriculum on loops, memory, tools, harness, orchestration, observability, and evals — see Agent Engineering.
2026 skills
Claude Code, skills files, loop engineering → AI Engineering 2026
The agentic stack¶
flowchart TB
subgraph Runtime["Harness & Runtime"]
Loop[Agent Loop]
State[State / Memory]
Term[Termination]
Perm[Permissions]
end
subgraph Capabilities["Capabilities"]
LLM[LLM Reasoning]
Tools[Tools & MCP]
RAG[RAG / Retrieval]
end
subgraph Scale["Scale Out"]
Orch[Orchestrator]
Workers[Worker Agents]
Handoff[Handoffs]
end
User --> Runtime
Runtime --> Capabilities
Runtime --> Scale
Runtime --> Obs[Observability]
Learning path (agentic track)¶
| Step | Course | What you'll learn |
|---|---|---|
| 1 | Course 07 · AI Agents | Agent loop, ReAct, tool use, frameworks |
| 2 | Course 08 · Harness & Tools | Runtime primitives, MCP, safety, tracing |
| 3 | Course 09 · Multi-Agent | Orchestration, coordination, patterns |
| 4 | Course 06 · RAG | Retrieval-driven agents |
| 5 | Course 16 · Capstones | End-to-end agent projects |
Core concepts¶
| Concept | Handbook | OSS inspiration |
|---|---|---|
| Agent loop | Course 07 · Intro to agents | Microsoft AI Agents for Beginners |
| Harness | Course 08 | Awesome Harness Engineering |
| Tools & MCP | Course 08 · Tools & MCP | Model Context Protocol |
| Orchestration | Course 09 | Agents Towards Production |
| Workflow vs agent | Course 07 · Workflow vs agent | LangGraph docs |
When to use what¶
| Problem | Pattern | Course |
|---|---|---|
| Single task with tools | ReAct agent | 07 |
| Long-running coding agent | Harness + sandbox | 08 |
| Research across sources | Multi-agent + RAG | 09, 06 |
| Deterministic pipeline | Workflow (not agent) | 07 |
| Customer support | Orchestrator + specialists | 09 |
Visual references¶
- The Illustrated Transformer — attention foundation for reasoning models
- How GPT-3 Works — token prediction → tool selection analogy
Related¶
- Evals & Observability — measure and debug agent runs
- Topic Map — find any concept
- Glossary — agent, harness, MCP, trajectory eval