Course · 8 chapters
Ingénierie des agents
Build production AI agents that use tools, remember across sessions, and recover from failures. Hands-on with MCP, Agent Skills, and agentic RAG, 8 chapters for engineers.
What you'll be able to do
- Build your first AI agent from scratch
- Apply ReAct loops and agentic patterns
- Design agent memory that persists
- Connect Claude to tools with MCP
- Ship agentic RAG that self-corrects
- Make agents resilient to failure
What's inside
- 1Agent Engineering : par où commencer
Une orientation de 12 minutes sur le parcours Agent Engineering en 7 chapitres — le chapitre de base, puis les primitives dont tu as besoin avant de choisir un archétype d'agent.
- 2Ton premier agent IA
Boucle, outil, mémoire — construis le plus petit agent possible et observe exactement ce qui le fait fonctionner.
- 3Patrons agentiques
Utilisation d'outils, boucles ReAct, orchestration multi-agents et garde-fous pour des systèmes d'agents fiables.
- 4Mémoire et gestion de l'état des agents
Conçois des systèmes de mémoire qui offrent aux agents IA continuité, personnalisation et capacité d'apprentissage entre les sessions.
- 5MCP avec Claude Code
Connecte Claude à tes outils du quotidien grâce au Model Context Protocol.
- 6Maîtriser les Agent Skills
Construis des paquets d'instructions réutilisables qui enseignent tes workflows aux agents IA — tu n'expliques qu'une seule fois et tu en profites à chaque utilisation.
- 7RAG Agentique
Construis des agents de recherche qui planifient leurs requêtes, routent entre les sources et s'auto-corrigent — le pattern d'architecture n°1 en production IA.
- 8Défaillances et récupération des agents
Construis des agents qui survivent aux échecs d'outils, aux boucles incontrôlées et aux erreurs en cascade — avec des tentatives de reprise, des disjoncteurs et une escalade humaine.
Frequently asked questions
- What will I learn in the Agent Engineering course?
- You learn to build production AI agents from the primitives up: tool use, memory and state, the Model Context Protocol (MCP), Agent Skills, agentic RAG, and failure recovery with retry and circuit breakers. It covers 8 chapters across roughly 2.7 hours.
- Who is this course for?
- It is built for software engineers and AI engineers at a practitioner level who want to design and ship reliable agent systems, not just prototypes. The path assumes you can read and write code.
- Do I need coding experience to take this course?
- Yes. This is an engineering-focused, practitioner-level path, so you should be comfortable with code and building software. It walks through hands-on agent construction rather than concepts alone.
- How long is the course and is there a certificate?
- The path has 8 chapters totaling about 162 minutes (roughly 2.7 hours). After you complete all chapters you earn a certificate from AI Academy by Anthropos.
- Is the Agent Engineering course free?
- No. Agent Engineering is a paid path included with an AI Academy by Anthropos subscription. It sits in the AI for Engineers track alongside other AI engineering content.
Earn a certificate
Complete all chapters to receive your certificate of completion.