Skill
Agent architecture
A loop: observe, think, act, observe. Keep it small enough to debug.
Do these first
What you will be able to do
- Implement a tool-using loop in <150 lines
- Log every tool call
- Stop on max steps
Curated resources
Start with the pick, then branch out
Scores are LearnPath editorial opinions (clarity, cost, freshness, project density) — not an objective ranking. Every YouTube item is embedded officially after a click-to-load facade.
Project-based
LearnPath pickinteractiveintermediate
Hugging Face agents course
Hugging Face6h
Free structured path. Prefer building a tiny agent over collecting frameworks.
Score 88
OpenProject-based
Alternativedocumentationintermediate
Build an agent (LangChain conceptual docs)
LangChain1h
Libraries churn. Learn the ideas (tools, memory, control loop), not the import paths.
Score 82
Open
2h 14mProject-basedProject-orientedyoutube videointermediate
Guide to Agentic AI – Build a Python Coding Agent with Gemini
freeCodeCamp.org2h 14m
Two hours building a coding agent with the Gemini API — file editing tools, sandboxing, iteration loops — the architecture behind tools like code assistants. A strong project after the MCP and tool-calling resources.
Score 81
Watch
1h 4mBest for beginnersOfficial docsyoutube coursebeginner
Full Course (Lessons 1-10) AI Agents for Beginners
Microsoft Developer1h 4m
Microsoft's 'AI Agents for Beginners' lessons 1–10 in a single hour: what an agent is, tool use, planning, memory, evaluation — vendor-official teaching material with small code samples. The best structured first contact with agents.
Score 79
Watch
24h 16mLearnPath pickMore curatedyoutube courseintermediate
Agentic AI – Complete Course for Beginners
freeCodeCamp.org24h 16m
Twenty-four hours published July 2026 — the most current and most complete agentic AI course verified: agent loops, tool calling, memory, planning, multi-agent patterns and production concerns, each with implementations. This is the spine of the agentic roadmap.
Score 89
WatchApply it
Projects