AI for developers
AI agents: architectures, frameworks and going to production
Design, develop and operate agents and multi-agent systems with today's tools.
- Duration
- 14 h · 2 days
- Level
- Advanced
- Format
- On site · Remote · Blended
- Max. group size
- 12 participants
Overview
This program takes stock of AI agents and how to implement them: taxonomy, reasoning models, function calling, classic and agentic RAG, MCP and A2A protocols, orchestration frameworks, multi-agent patterns, security and guardrails, observability, evaluation and deployment. Participants leave with a reference architecture and a working agent.
Who this program is for
- Developers and architects
- Data and AI engineers
- Technical managers preparing agent projects
Prerequisites
- Command of a programming language, ideally Python
- Basic knowledge of LLMs (prompts, temperature, context window)
- Understanding of REST APIs and application architectures
- Comfortable with the command line
Learning objectives
- Describe the state of the art of AI agents
- Design an agent's architecture: model, RAG, tools, memory, protocols
- Develop simple then multi-agent agents
- Integrate agents into an IT system via APIs, business tools, MCP and A2A
- Apply security, guardrails and governance
- Set up observability, evaluation and deployment
Detailed curriculum
4 modules · 14 h · 2 days
Day 1
Fundamentals and overview
30% theory · 70% practice
- Agent taxonomy and levels of autonomy
- Market trends
- Reasoning models and function calling
- Classic RAG and agentic RAG
- MCP and A2A protocols
🛠 Hands-on workshops
Scope an agent1 h
Define an agent's scope, tools and guardrails.
✅ Deliverable : A design sheet
Build a first agent
30% theory · 70% practice
- Perception, reasoning, action loop
- Defining reliable tools
- Memory and context
- Error handling
- Logging
🛠 Hands-on workshops
Tool-using agent2 h
Develop an agent that uses several tools.
✅ Deliverable : A working agent
Day 2
Frameworks and multi-agent
30% theory · 70% practice
- Overview of agent frameworks
- Graph-based orchestration
- Multi-agent patterns: supervisor, pipeline, debate
- Choosing a framework
- Interoperability
🛠 Hands-on workshops
Multi-agent system2 h
Orchestrate several agents on an enterprise case.
✅ Deliverable : A multi-agent prototype
Security, operations and deployment
30% theory · 70% practice
- Guardrails, permissions and human validation
- Prompt injection and tool abuse
- Observability and traces
- Evaluation and deployment pipeline
- Use cases and outlook
🛠 Hands-on workshops
Go-live plan1 h
Write the production rollout plan for an agent.
✅ Deliverable : A production plan
Tools and technologies
API AnthropicAPI OpenAIDockerLangChainMCPPythonVisual Studio Code
Teaching methods
- Alternating short lectures and hands-on work
- Live demonstrations by the trainer
- Capstone project built throughout the modules
- Group corrections and discussion of best practices
Assessment
- Hands-on exercises corrected in each module
- Mini-project completed at the end of the course
- Knowledge-check quiz
Required equipment
- A computer with an internet connection
- Python 3, Git, Docker (recommended) and model API keys (provided or personal)
Resources provided
- Course materials and repository of corrected exercises
Career outcomes
- An operational agent with tools
- A reference architecture
- A production-readiness grid
Related video courses
My video courses that go with this program, to get ready for the training or to review afterwards. Free with an account.
