O'Reilly Live Learning Course | 4 Hours | AI Agents -- From Landscape to Code
Build AI agents across the full spectrum -- from declarative M365 Copilot agents and low-code Copilot Studio bots to code-first Python agents with Azure AI Foundry, LangGraph, and FastMCP.
| Hour | Theme | Key Platforms & Tools |
|---|---|---|
| 1 | What "Agent" Means | M365 Copilot agents (custom GPTs), declarative agents, agent landscape taxonomy |
| 2 | Low-Code Agents | Copilot Studio, Antigrav, Claude Code / GitHub Copilot agents |
| 3 | Code-First Agents | Azure AI Foundry agents, Python, LangGraph, FastMCP |
| 4 | TBD | Best practices, MCP deep dive, future trends (finalize before delivery) |
- Microsoft 365 account (Business or Enterprise) -- start free trial
- Copilot Studio access -- start free trial
- Power Automate Premium -- start 90-day trial
- Microsoft Teams desktop app
- Basic familiarity with Microsoft 365 apps
Hour 3 additionally requires:
- Python 3.11+ -- https://www.python.org/downloads/
- Azure subscription (free tier OK) -- https://azure.microsoft.com/en-us/free
Optional trials for extended scenarios:
| Resource | URL | Duration |
|---|---|---|
| Power Apps Developer Plan | https://www.microsoft.com/power-platform/products/power-apps/free | No limit |
| Azure Free Account | https://azure.microsoft.com/en-us/free | 30 days ($200 credit) |
| Power Platform Trials Hub | https://www.microsoft.com/en-us/power-platform/try-free | Various |
agents-pro/
├── README.md # This file
├── CLAUDE.md # Claude Code instructions
├── CODE_OF_CONDUCT.md # Community guidelines
├── contributing.md # Contribution guide
├── SECURITY.md # Security policy
├── LICENSE # MIT License
├── markdownlint.json # Markdown linting config
│
├── .github/ # GitHub Copilot configurations
│ ├── agents/ # Copilot agent definitions
│ ├── prompts/ # Copilot prompt templates
│ └── instructions/ # Copilot coding instructions
│
├── agents-playground/ # Sandbox workspace + experiments
├── docs/ # Course plan, WAF notes, research
├── images/ # Course images and assets
├── src/ # Canonical agent implementations
│ ├── claude-agent/ # Claude Code agent materials
│ ├── copilot-studio/ # Copilot Studio topic assets
│ ├── copilot-studio-agent/ # Copilot Studio agent scaffolds
│ ├── foundry-agent/ # Azure AI Foundry code-first agent
│ └── oreilly-agent-mvp/ # O'Reilly agent MVP workspace
└── tests/ # (currently empty)
-
Clone the repository
git clone https://github.com/timothywarner-org/agents-pro.git
-
Hour 1 -- Agent Landscape: Use the teaching guides in
agents-playground/docs/(seehour-1-teaching-guide.md) -
Hour 2 -- Low-Code Agents: Build from
src/copilot-studio-agent/and reference assets insrc/copilot-studio/oragents-playground/copilot-studio/ -
Hour 3 -- Code-First Agents: Start with
src/foundry-agent/and the MVP workspace insrc/oreilly-agent-mvp/ -
Hour 4 -- TBD: Track the draft plan in
docs/course-plan-february-2026.md
This repo tracks PowerPoint decks with Git LFS. Install Git LFS before cloning or pushing PPTX updates.
git lfs install
git lfs pullTheme: Orientation and taxonomy
- Agent landscape -- what counts as an "agent" and what does not
- M365 Copilot agents (custom GPTs) and declarative agents
- Where Copilot Studio, Azure AI Foundry, and open-source frameworks fit
- Hands-on: explore a declarative agent inside M365 Copilot
Skills: Agent taxonomy, M365 Copilot configuration, declarative agent basics
Theme: Build agents without writing code
- Copilot Studio -- topics, triggers, knowledge sources, generative answers
- Antigrav low-code agent builder
- Claude Code and GitHub Copilot as agentic coding assistants
- Hands-on: build and publish a Copilot Studio agent to Teams
Skills: Copilot Studio authoring, Power Automate integration, low-code agent patterns
Theme: Python-based agents with full control
- Azure AI Foundry agents -- deploy and orchestrate from code
- LangGraph -- stateful, multi-step agent graphs
- FastMCP -- lightweight Model Context Protocol servers
- Hands-on: wire an Azure AI Foundry agent to a LangGraph workflow with MCP tools
Skills: Python agent development, Azure AI Foundry SDK, LangGraph, FastMCP, MCP
Theme: To be finalized before delivery
- Best practices for production agent deployments
- MCP deep dive -- protocol internals and tool authoring
- Future trends in the agent ecosystem
- Wrap-up exercise: identify three processes to automate next
- Copilot Studio Documentation
- Topic Authoring Best Practices
- Agent Flows Overview
- Copilot Studio VS Code Extension (Preview)
- Azure AI Foundry
- LangGraph Documentation
- FastMCP
- Model Context Protocol
- Microsoft Agent Framework | GitHub
- Microsoft 365 Agents SDK
- MS-4004: M365 Copilot Use Cases
- AI-102: Azure AI Engineer Associate
- Copilot & Agent Administration Fundamentals
| Time | Activity |
|---|---|
| 0:00 - 0:50 | Hour 1: What "Agent" Means -- taxonomy, M365 Copilot agents, declarative agents |
| 0:50 - 1:00 | Q&A + Break |
| 1:00 - 1:50 | Hour 2: Low-Code Agents -- Copilot Studio, Antigrav, Claude Code / GitHub Copilot |
| 1:50 - 2:00 | Q&A + Break |
| 2:00 - 2:50 | Hour 3: Code-First Agents -- Azure AI Foundry, Python, LangGraph, FastMCP |
| 2:50 - 3:00 | Q&A + Break |
| 3:00 - 3:50 | Hour 4: TBD -- best practices, MCP deep dive, future trends |
| 3:50 - 4:00 | Wrap-up, resources, next steps |
Tim Warner -- Microsoft MVP (Azure AI and Cloud/Datacenter Management), Microsoft Certified Trainer
MIT License - See LICENSE for details.
Questions? Open an issue or reach out via the course chat during live sessions.