Building your own Agents with Copilot Studio

Part 1 - Copilot Studio Fundamentals

Copilot Studio and the Power Platform

This module introduces the various Microsoft Copilot products available today. You will discover how Copilot Studio leverages Power Platform features to support enterprise-grade governance and lifecycle management.Finally, you will get familiar with the three distinct harnesses that determine how your agent reasons, what it can do out of the box, and how it's billed.

• The Microsoft Copilot Landscape
• Understanding Agents
• The Power Platform and Microsoft Dataverse
• Understanding Environments & Solutions
• Introduction to Microsoft Copilot Studio
• The GitHub Copilot Harness vs. The Standard Harness
• Comparing Harness Capabilities and Performance
• Licensing and Copilot Credits
• Estimating and Monitoring your Consumption
• LAB: Prepare Your Environment for Agent Creation

Creating Declarative Agents

Declarative agents are powered by Microsoft 365 Copilot. This chapter explains how to build these declarative agents outside Copilot Studio.

• What are Microsoft 365 Copilot Declarative Agents?
• Creating Declarative Agents with Agent Builder
• Publishing and Sharing Declarative Agents
• Building Agents in SharePoint and Teams
• LAB: Create Your First Declarative Agent

Part 2 - GitHub Copilot Harness

Building Your First Agent with GitHub Copilot Harness

This module introduces the new Agent Designer. How you describe an agent's behavior, how the Build, Preview, Evaluate and Monitor tabs work together, and how the harness's enhanced orchestration reasons through a request.

• The Agent Designer
• Build, Preview, Evaluate and Monitor Tabs
• Describing an Agent's Instructions
• Testing Your Agent in the Preview Tab
• Understanding the Enhanced Orchestration Runtime
• LAB: Create Your First Agent with GitHub Copilot Harness

Knowledge & Tools

Ground your agent in real organizational knowledge and give it the ability to take action. This module covers connecting knowledge sources and adding tools, including the native ability to create and edit Word, Excel, PowerPoint and PDF files.

• Adding Knowledge Sources: Websites, Documents and SharePoint
• Native File Creation and Editing (Word, Excel, PowerPoint, PDF)
• Adding Tools: Connectors, APIs and MCP Servers
• Connected Agents
• LAB: Enhancing Agents with Knowledge and Tools

Skills & Memory

Skills and memory are two capabilities unique to the GitHub Copilot harness that let an agent go beyond a single conversation turn. This module explains what a skill is, how it differs from a tool, and how memory lets an agent retain useful context across tasks and time.

• What is a Skill
• Adding and Configuring Skills
• Introduction to Agent Memory
• What Persists Across Turns and Tasks
• Designing Agents for Longer-Running Work
• LAB: Adding Skills and Memory to Your Agent

Publishing Agents

Once your agent is developed, it's essential to thoroughly test its performance before making it available to users. After testing, you can publish your agent across multiple platforms and monitor how it performs through detailed analytics. This chapter walks you through the full process: testing your agent, configuring authentication, sharing it with others, publishing to various channels, and analyzing its overall effectiveness.

• Testing Agent Performance
• Setting up Authentication for Agents
• Authentication Modes
• Monitoring Agent Quality and Performance
• Publishing Agents to Channels
• Sharing Agents with Users and Teams
• Tracking Copilot Credit Consumption After Publishing
• LAB: Publish Your Agent to a website

Workflows

Workflows are the answer to deterministic, repeatable automation. This module shows how workflows differ from Agents, and how agents and workflows call on each other to combine adaptability with predictability.

• Workflows vs Agents
• Building a Workflow
• Adding an Agent Node Inside a Workflow
• Using a Workflow as a Tool Inside an Agent
• Workflow Triggers
• Testing a Workflow
• Human in the Loop for Governance
• The Spectrum: Deterministic vs Generative
• The Hybrid Pattern: Adaptive Where You Want It, Controlled Where You Need It
• LAB: Create a Workflow and Use it Inside an Agent

Part 3 - Standard Harness

Building Your First Agent with Standard Harness

Learn how to build your first agent using the Standard harness in Copilot Studio. This chapter introduces the fundamentals of agent creation and orchestration modes, framed against what you already know from the GitHub Copilot harness.

• Creating a New Agent
• Setting Agent Name, Description and Settings
• Understanding Generative vs Classic Orchestration
• How This Compares to the GitHub Copilot Harness
• LAB: Create Your First Agent

Knowledge & Tools

Ground your agent in organizational knowledge and extend it with tools. This chapter covers connecting knowledge sources and defining connector and MCP tools.

• Overview of Knowledge Sources
• Adding Public Website, Document and Dataverse Knowledge
• SharePoint and OneDrive Knowledge
• Adding Knowledge via Copilot Connectors
• Creating Connector Tools
• Integrating MCP Servers as Tools
• Invoking Tools
• LAB: Enhancing Agents with Tools and Knowledge

Agent Flows & Topics

Topics and Agent Flows are the Standard harness's two deterministic building blocks. This chapter teaches both together: Topics for conversation-level determinism, Agent Flows for workflow-level determinism.

• Creating and Editing Topics
• Messages, Questions, Branching, Adaptive Cards
• Creating and Using Agents Flows in Copilot Studio
• Orchestrating Topics ang Agent Flows Using Generative AI
• LAB: Working with Topics & Agent Flows

Autonomous Agents

Autonomous agents are capable of executing actions automatically based on events or triggers, without human intervention. This module teaches how to configure Standard harness agents that respond to real-world signals such as incoming emails, new data entries, or scheduled events.

• Introduction to Autonomous Agents
• Adding Event Triggers in Copilot Studio
• Event Triggers and Power Automate
• Publishing Agents with Event Triggers
• LAB: Building an Autonomous Agent

Publishing Standard Harness Agents

Once your agent is developed, it's essential to thoroughly test its performance before making it available to users. After testing, you can publish your agent across multiple platforms and monitor how it performs through detailed analytics. This chapter walks you through the full process: testing your agent, configuring authentication, sharing it with others, publishing to various channels, and analyzing its overall effectiveness.

• Testing Agent Performance
• Setting up Authentication for Agents
• Authentication Modes
• Monitoring Agent Quality and Performance
• Publishing Agents to Channels
• Sharing Agents with Users and Teams
• Tracking Copilot Credit Consumption After Publishing
• LAB: Publish Your Agent to a channel