Build AI agents from the ground up
Start with a simple prompt-based agent, then improve its instructions and behavior.
In-person workshop / Built for builders
From AI Agents to Agentic Workflows
Turn AI agent concepts into practical skills—one small, testable step at a time.
A hands-on workshop with GitHub Copilot & Microsoft Foundry
Reserve your placeAt Arrow Oslo. Bring your curiosity.
Save your seat
Register for the workshop at Arrow Oslo.
Preview mode: registration opens once the Eloqua connection and event details are configured.
WHEN19 November 202609:00–15:00 CET (Oslo, UTC+1)
WHEREArrow OsloArrow, Innspurten 1A, 0663 Oslo
FORMATHands-on workshop6 hours
IN THE TOOLKITGitHub Copilot +
Microsoft Foundry
Microsoft Foundry
The workshop
Two agents.
One connected experience.
This hands-on workshop takes you from a basic AI agent to a complete multi-agent experience, with GitHub Copilot as your guide and Microsoft Foundry as the place where your agents run. You’ll build two focused agents, connect them to useful tools, teach them reusable procedures, and give them memory. Then you’ll bring them together in a hosted workflow that assembles a sourced travel itinerary.
You won’t just follow a recipe: at every stage, you’ll make a change, test it, inspect the evidence, and explain what happened. Along the way, you’ll also practice using the GitHub Copilot app to understand code, review changes, inspect tool calls, and resume work across sessions.
Build understanding
What you’ll learn
From a first agent to a tested workflow.
Work with conversations and memory
Understand how managed conversation history differs from information an agent can remember across conversations.
Connect agents to tools
Use MCP servers to let agents call specific operations, such as retrieving a weather forecast or finding cultural attractions.
Guide tool use with skills
Give an agent a reusable procedure for deciding how to approach a task—without confusing that guidance with access to tools.
Test and release with confidence
Check behavior at each stage, review traces and results, and release agents only after they pass the workshop’s tests.
Coordinate multiple agents
Define what information specialists should return, then use a hosted workflow to invoke them and combine their results.
Use GitHub Copilot more effectively
Practice asking for explanations, making focused changes, reviewing diffs, inspecting tool calls, and resuming sessions.
Bring your curiosity
Who it’s for
This workshop is designed for developers, technical professionals, and curious builders who want to understand how AI agents work in practice—not just what they are. It’s suitable if you’re new to agent development, as well as if you’ve experimented with prompts and want a more structured way to explore tools, memory, testing, and orchestration.
You don’t need prior experience with Microsoft Foundry, MCP, or multi-agent systems. The coach introduces new terms when they’re needed and guides you through one manageable change at a time. VS Code is not required; the main learning experience takes place in the GitHub Copilot app.
One manageable change at a time
How it works
Two specialists. A hosted workflow. Your capstone.
Get set up
Use your assigned Microsoft Foundry Project and choose a compatible model deployment for your exercises.
Build the Weather Research Agent
Create a baseline agent, refine its instructions, add managed conversations, connect a weather tool, and work up to skills, memory, and release.
Build the Cultural Travel Agent
Repeat the progression with a separate specialist focused on cultural attractions and its own tools and evidence.
Bring the agents together
Once both specialists pass independently, build and test a hosted workflow that calls them and assembles a sourced itinerary.
Apply your learning
Choose a capstone scenario—such as IT support, documentation research, fictional travel planning, or an approved idea of your own—and use the same principles to build a solution.
Each lesson follows a practical cycle: predict, change, test, explain, and checkpoint. You’ll see what changed, verify the result, and build understanding before moving on.
Before you arrive
Come ready to build.
Required access: Access to an active Azure subscription for Microsoft Foundry, and access to GitHub.
Participants choose their own compatible model deployment; availability, quota, and possible costs depend on their Foundry project.
Your next step
Put agents
into practice.
You’ll leave with more than a working example: you’ll have practiced a repeatable approach for building, testing, releasing, and coordinating AI agents.
- Two focused agents with tools and memory
- A tested, hosted multi-agent workflow
- A capstone scenario to apply your learning
Predict. Change. Test.
Explain. Checkpoint.
THE AGENTIC WORKFLOW01 — 03
01 / WEATHERWeather Research Agent
02 / CULTURECultural Travel AgentTools · skills · memory
03 / HOSTED WORKFLOWA sourced travel itinerary.
+ HUMAN JUDGMENT, THROUGHOUT +
Good to know
A few details.
Do I need experience with agents?
You don’t need prior experience with Microsoft Foundry, MCP, or multi-agent systems. The coach introduces new terms when they’re needed and guides you through one manageable change at a time.
Do I need VS Code?
No. VS Code is not required; the main learning experience takes place in the GitHub Copilot app.
What access do I need?
Access to an active Azure subscription for Microsoft Foundry, and access to GitHub.
Where is the workshop?
This is an in-person workshop at Arrow Oslo, not an online event. Arrow, Innspurten 1A, 0663 Oslo