Microsoft Copilot Studio Labs

Using Multi-Agent in Copilot Studio

⏱️ 30 minutes 📊 Level 300 🏷️ Local
📄 Download PDF

Help makers understand how to leverage multi-agent configurations including child and connected agents.

🧭 Lab Details

Level Persona Duration Purpose
300 Maker 30 minutes After completing this lab, participants will be able to leverage multi-agent configurations with Copilot Studio. This lab teaches the concepts of multi-agent and shows the difference of connected and child agent configurations.

📚 Table of Contents

🤔 Why This Matters

Makers and Architects - Want to know how to appropriately leverage multi-agent configurations and when to use connected vs child agents to optimize how your agent orchestrates and is built for scale.

Think of a complex agent that might have many tools and knowledge sets that should only be used when appropriate:

  • Without Multi-Agent: Makers are heavily limited in the number of orchestratable topics and tools they have at their disposal.
  • With Multi-Agent: Makers are able to logically group their tools, topics, and knowledge into contextualized agents that allow for additional instructions and descriptions to be used to guide their orchestration on the best way to find the data it is looking to provide to the user.

Common challenges solved by this lab:

  • "Abusing connected agents"
  • "Not knowing if something should be a connected agent or child agent"
  • "Don't understand the capabilities and rationale of child agents"
  • "Not classifying their agents' capabilities and putting excessive instructions in the overview instructions to compensate"

Become a multi-agent expert - no need to continue to wonder about how to work around limitations of tools and topics that can be orchestrated or have the wrong tool being called at the wrong time.

🌐 Introduction

Welcome to the Copilot Studio Multi‑Agent Lab. In this hands‑on session, you’ll learn how to design, orchestrate, and extend intelligent copilots using multi‑agent patterns, including child agents, connected agents, and coordinated task delegation. This lab is designed to help you understand how complex workflows can be broken into smaller, specialized agents that collaborate to solve problems—mirroring modern enterprise AI architectures. You’ll build and configure agents, define how they communicate, and connect them to real data and actions. By the end, you’ll have a working multi‑agent copilot capable of distributing tasks, retrieving information from external systems, and chaining logic across multiple agent types.

What You Will Learn

  • How multi‑agent patterns work in Copilot Studio
  • Creating and configuring child agents to handle specialized tasks
  • Chaining logic so a parent agent delegates tasks to child agents
  • Building connected agents that access data sources and external systems
  • Designing agent‑to‑agent communication flows
  • Using dynamic chaining to route user intent to the right agent
  • Implementing branching logic for conditional task routing
  • Best practices for agent design, naming, and responsibility boundaries
  • Testing multi‑agent scenarios end to end in Copilot Studio
  • Troubleshooting common multi‑agent issues

🎓 Core Concepts Overview

Concept Why it matters
Child Agents Allows a maker to group a common set of tools and knowledge including instructions into a logical group
Connected Agents Allows a maker to connect an external agent to another agent in Copilot Studio
Orchestration Allows a maker to let the agent decide what tools and agents are best to solve the user's request

📄 Documentation and Additional Training Links

✅ Prerequisites

  • Access to Microsoft Copilot Studio
  • Dataverse search enabled in environment
  • Access to Dataverse unbound action connector
  • Sample data loaded into Dataverse tables
  • Access to Account / Contact table in environment
  • Access to modify views and create search indexes on Account / Contact table

🎯 Summary of Targets

In this lab, you'll build a parent agent that will leverage both connected agents and child agents. By the end of the lab, you will:

  • Create a parent agent for a Sales Associate
  • Configure a child agent that gets information about products
  • Connect to a connected agent that gets information about Accounts and Contacts
  • Use generative orchestration to select the correct agent and tools to execute the request

🧩 Use Cases Covered

Step Use Case Value added Effort
1 Create Sales Assistant Agent Creates parent agent to help sales associate find information they need 8 min
2 Create Product Information Child Agents Creates child agents to provide market-specific product information 12 min
3 Connect to Account and Contact Information Agent Connect existing Copilot Studio agent that provides account and contact information 10 min
4 Connect a Fabric Data Agent (Optional) Extend your multi-agent solution with Fabric Data Agent for e-commerce analytics Optional

🛠️ Instructions by Use Case

🧱 Use Case #1: Create Sales Assistant Agent

Creates parent agent to help sales associate find information they need

Use case Value added Estimated effort
Create Sales Assistant Agent Creates parent agent to help sales associate find information they need 8 minutes

Summary of tasks

In this section, you'll learn how to create a parent agent to use as the base agent for a Sales Associate Assistant.

Scenario: Your Sales Team needs to quickly answer questions and help customers with information on your products and with finding account and contact information.

Objective

Create the parent agent base that will be needed to host all your tools, agents, and knowledge for sales associates.

Step-by-step instructions

Create Sales Associate Assistant

  1. Go to the Copilot Studio home page at https://copilotstudio.microsoft.com.

  2. Check the Environment in the upper right corner to make sure you are in your "DEV - user name" environment.

  3. On the Home page in the description of the agent you want to build, enter the following but do not select enter yet:

     Agent that assists sales associates with getting product information and lookup account and contact information in the sales system.

    [!IMPORTANT] Do not Select Enter or submit this text yet as we need to do the agent settings before we do that.

Home Creation Text

  1. Select the Gear on the description input area and edit the schema name to include salesassistant and then Select Update.

    Agent Settings

    [!TIP] Putting an agent into a dedicated solution and changing the schema name is always helpful for application lifecycle management in the future.

  2. Select Enter or the Arrow button on the right side of the description field.

  3. Once your agent is fully provisioned, review the instructions and tool or knowledge recommendations provided by the Copilot Studio provisioning process.

Prevent Hallucinations

  1. To make sure that our agent doesn't use model knowledge, we want to turn off some features. Select Settings in the upper right corner.

  2. Scroll to the bottom of the list of settings on the Generative AI settings screen and turn Off both Use general knowledge and Use information from the Web.

    Turn off Model and Web Knowledge

  3. Select Save and then select the X in the upper right corner of the settings page to close it.

    [!TIP] Turning off model and web knowledge helps to keep your agent from hallucinating information that the model might know about our product catalog or other items that might confuse it. This is best practice if you want it to keep to just the content you provide.

Collect Sales Associate's Market

  1. Select Topics in the top navigation menu.

  2. Select Add a Topic and then select From blank

    Create blank topic

  3. Select the box that says The agent chooses and then select the Double Arrows to show the Change Trigger menu, then select It's redirected to

    Set Trigger on Topic

  4. Change the name of the Topic to Select Market by selecting on the name in the upper left navigation.

    Set Topic Name

  5. Add a node after the Trigger by selecting the + and select Ask a question

    Add Question

  6. Enter What market are you wanting information in? in the message area.

  7. Select New option and add US as the option, and then repeat to add UK as another option.

    Create Question and Options

  8. Under Save user response as, select Var1 and change the Variable name to Market.

  9. On the Variable properties pane, select Global as the Usage of the Variable and select External sources can receive value.

  10. Enter The market that the sales associate is requesting information about. in the Variable description, near the bottom of the Variable properties.

    Variable Settings

  11. Delete all the Condition nodes that were automatically added below the Question node by selecting … and then selecting Delete on each

    Delete Conditions

  12. Select Save to save the topic

  13. Select the V next to the name of the topic "Select Market" and select Conversation Start to go to the Conversation Start Topic

    Go To Conversation Start

  14. Add a new node at the end of the topic by selecting the + and then selecting Topic Management > Go to another topic > Select Market

Add Node to Conversation Start

  1. Select Save

  2. Open the Test chat by Selecting Test in the upper right corner.

  3. Select the + in the Test chat to start a new test session and verify that you now are asked which market at the beginning of each new chat.

    Test Chat Market Selection

🏅 Congratulations! You've completed Use Case #1!

Test your understanding

Key takeaways:

  • Schema naming matters – Using a descriptive schema name helps with application lifecycle management and finding your agent later.
  • Disable model knowledge early – Turning off general knowledge and web search prevents hallucinations and ensures the agent only uses content you provide.
  • Global variables enable cross-agent context – By making the Market variable global with external send capability, child agents can use it for conditional logic.

Lessons learned & troubleshooting tips:

  • If the market question doesn't appear, verify the conversation start topic redirects to your Select Market topic.
  • If the variable isn't available in child agents, confirm it is set to Global with External sources can receive value enabled.

🔄 Use Case #2: Create Product Information Child Agents

Create child agents that allow the user, based on their market, to get the right product information.

Use case Value added Estimated effort
Create Product Information Child Agents Group common tools and knowledge into logical groups for orchestration optimization 12 minutes

Summary of tasks

In this section, you'll learn how to create child agents to logically group knowledge and tools allowing them to be orchestrated based on a condition.

Scenario: You want to orchestrate to knowledge about your products based upon the market that the user selects.

Create Child Agent for US product information

  1. In the Sales Associate Assistant, select Agents on the top navigation menu.

  2. Select Add on the Agents page.

  3. Select New child agent.

Select Child Agent

  1. Set Agent Name to US Product Information Agent.

  2. Set the Description to This agent provides details about the Surface products offered in the United States

  3. Expand Advanced and set the Condition to the Global.Market variable is equal to US

    Set Child Agent Details

    [!TIP] You can use a formula or variables to make conditions for when this agent should be available. If you don't see the Global.Market variable that probably means you didn't mark it Global in the topic.

  4. Enter the following in the Instructions

     This agent should only answer questions about Microsoft Surface products in regards to the US-based products.  It should never provide details or stats on any products not offered or details about models offered outside of the United States.  You should only talk about Microsoft Surface products and not discuss any products that are not Microsoft Surface products even if they are made by Microsoft.

Add instructions

  1. Add Knowledge to the Child Agent by Selecting Add in the Knowledge section of the Child Agent configuration

  2. Select Public Websites

  3. Enter microsoft.com in Public website link and Select Add

  4. Select Add to agent

  5. Select Save to save your child agent

Create Child Agent for UK product information

  1. In the Sales Associate Assistant, select Agents on the top navigation menu

  2. Select Add on the Agents page

  3. Select New child agent

Select Child Agent

  1. Set Agent Name to UK Product Information Agent

  2. Set the Description to This agent provides details about the Surface products offered in the United Kingdom

  3. Expand Advanced and set the Condition to the Global.Market variable is equal to UK

Set Child Agent Details UK

> [!TIP]
> You can use a formula or variables to make conditions for when this agent should be available.

  1. Enter the following in the Instructions:

     This agent should only answer questions about Microsoft Surface products in regards to the UK-based products.  It should never provide details or stats on any products not offered or details about models offered outside of the United Kingdom.  You should only talk about Microsoft Surface products and not discuss any products that are not Microsoft Surface products even if they are made by Microsoft.`

Add Child Agent Instructions UK

  1. Add Knowledge to the Child Agent by Selecting Add in the Knowledge section of the Child Agent configuration

  2. Select Public Websites

  3. Enter microsoft.com/en-gb in Public website link and Select Add

  4. Select Add to agent

  5. Select Save to save your child agent

  6. Open the Test chat by Selecting Test in the upper right-hand corner

  7. Select the + in the Test chat to start a new test session

  8. Select UK

  9. Enter What are the starting prices for the Surface laptop?

  10. Verify that the correct child agent was triggered and that your answer is appropriate for the UK market

Example Response UK question

  1. Reset the conversation and do again for the US market.

🏅 Congratulations! You've completed Use Case #2!

Test your understanding

Key takeaways:

  • Child agents group related capabilities – Grouping knowledge and tools into a child agent with its own instructions improves orchestration accuracy.
  • Conditions control availability – Using a variable condition on a child agent means the orchestrator only considers it when the condition is met, preventing irrelevant responses.
  • Instructions scope the agent's domain – Specific instructions telling the agent what NOT to answer are just as important as telling it what to answer.

Lessons learned & troubleshooting tips:

  • If the wrong child agent responds, verify the condition on the Advanced section matches the correct Market variable value.
  • If knowledge doesn't return results, confirm the Public Website knowledge source status shows a green check mark.

🧱 Use Case #3: Connect to Account and Contact Information Agent

Connect the Account and Contact Information agent to our parent Sales Associate Agent. Will need to ensure indexing is working in the environment ahead of testing.

Use case Value added Estimated effort
Connect to Account and Contact Information Agent Connect to an Account and Contact Information agent 10 minutes

Summary of tasks

In this section, you'll learn how to connect an existing Copilot Studio agent to your parent agent using the connected agents feature.

Scenario: Your Sales Team needs to quickly answer questions and help customers with information on your products and with finding account and contact information.

Objective

Connect existing agent to our Sales Associate Assistant agent to add ability to find account and contact information from the sales system.

Step-by-step instructions

Make sure that Dataverse Search is set to on

  1. In the upper right corner of Copilot Studio, Select the Gear icon

  2. Select Go to Power Platform admin center

  3. In the left navigation, select Manage then select Environments

  4. Select your environment from the list

  5. Select Settings in the top menu

  6. Expand Product and select Features

  7. In the Dataverse search section, ensure Dataverse search is turned On

  8. Select Save if you made any changes

Ensure that indexes are in place for our connected agent

> [!IMPORTANT]
> This is not required for a connected agent to work, but to make sure that the one we have pre-loaded for you will return results we must index a few tables in your environment.
  1. Go to the Power Apps maker portal https://make.powerapps.com

Open Power Apps

1 In the left menu select Tables

  1. Select the Account table from the list

  2. Select Views from the Data experiences section

  3. Select Quick Find Active Accounts option from the list of Views

  4. Select View Column to verify the following list of columns are in the view, you may have to scroll to see all of the included columns:
    • Address1: State or Providence
    • Address1: Postal Code
    • Address1: City
    • Annual Revenue
    • Currency
  5. To add any that are missing select + View column and then select the column from the list.

Account View

  1. Add the ability to search on certain fields by making sure that the following items are in the Find by on the bottom right. Select the Edit find table columns option to check:
    • Address1: State or Providence
    • Address1: Postal Code
    • Address1: City

Add Account Searchable Columns

  1. Select Edit find table columns to add any missing columns.

  2. Select Save and publish to update the index

    [!IMPORTANT] DO NOT navigate away until the save and publish is completed!

  3. Select Back in the upper left corner to go back to the Views list

  4. Select on Tables in the Views screen to go back to the list of Tables

Go to Tables

  1. Select Contact table from the list

  2. Select Views from the Data experiences section

  3. Select Quick Find Active Contacts option from the list of Views

  4. Select View Column to verify the following columns are in the view:

    • Anniversary
    • Birthday
    • Job Title
    • Marital Status

Column Add Contact

  1. Select + View column to add any that are missing.

  2. Select Save and publish to update the index

[!IMPORTANT] DO NOT navigate away until the save and publish is completed!

Test and Publish the Account and Contact Information Agent

  1. In the Copilot Studio tab in your browser, go to the Account Data Lookup Agent

  2. Open the Test chat by Selecting Test in the upper right-hand corner

  3. Enter What are the accounts in Texas?

  4. Verify that you get a response showing that the agent is working and the data is indexed

Test Account Lookup

  1. Select Settings in the upper right menu

  2. Make sure that the setting in the Generative AI menu in the Connected Agents section for Let other agents connect to and use this one is set to On

  3. Close the Settings menu by Selecting the X in the upper right-hand corner

  4. Select Publish and make sure that your agent publishes

[!IMPORTANT] You can't connect to an agent unless it is published

Connect the Account and Contact Information Agent

  1. In the Copilot Studio tab in your browser, go to the parent agent (Sales Info Assistant)

  2. Go to the Agents menu on the top navigation bar

  3. Select Add an agent

  4. Select Account Data Lookup Agent from the menu

Select Connected Agent

  1. Notice that the Description provides details on when to use this agent

  2. Make sure that the Pass conversation history to this agent is Checked

Validate Setting Connected Agent

  1. Select Add and configure

  2. Open the Test chat by Selecting Test in the upper right corner

  3. Select US in the test canvas to select that geography

  4. Enter What are all of the details of accounts in Texas? and send to agent

  5. See that the agent orchestrated to the connected agent which then called its child agent that then called numerous tools to answer the question.

Account Info Results

🏅 Congratulations! You've completed Use Case #3!

Test your understanding

Key takeaways:

  • Connected agents extend reach – Unlike child agents that live inside your agent, connected agents are independently published agents that your agent delegates to.
  • Publishing is required – A connected agent must be published and have "Let other agents connect to and use this one" enabled before it can be connected.
  • Conversation history improves context – Passing conversation history to the connected agent helps it understand the full context of the user's request.

Lessons learned & troubleshooting tips:

  • If the connected agent doesn't appear in the list, ensure it is published and the sharing setting is turned on.
  • If Dataverse search returns no results, verify the Quick Find views have the correct columns added and the view is saved and published.
  • If the agent can't find account data, check that Dataverse Search is enabled in the Power Platform admin center for your environment.

Challenge: Apply this to your own use case

  • What agents in your organization could benefit from being connected to a parent orchestrator?
  • How would you decide whether something should be a child agent or a connected agent?
  • Consider building a connected agent for a different data source (e.g., ServiceNow, Salesforce) and adding it to this same parent agent.

🔄 Use Case #4 (Optional): Connect a Fabric Data Agent

Extend your multi-agent Sales Associate Assistant by connecting a Microsoft Fabric Data Agent to provide e-commerce analytics capabilities through agent-to-agent communication.

Use case Value added Estimated effort
Connect a Fabric Data Agent Add Fabric data intelligence to your multi-agent solution for e-commerce analytics Optional (20 minutes)

[!NOTE] This use case is optional and does not count toward the 30-minute lab duration. It requires access to a Microsoft Fabric capacity and the Workshop Demo Workspace. Complete this if you finish the core lab early or want to explore advanced multi-agent patterns.

Summary of tasks

In this section, you'll learn how to create a Fabric Data Agent, optimize it with meta-prompts, and connect it as an additional agent to your Sales Associate Assistant.

Scenario: Your Sales Team wants to ask natural language questions about e-commerce performance data stored in Microsoft Fabric — such as revenue trends, top products, and customer segments — directly through the Sales Associate Assistant without needing SQL expertise.

Objective

Create a Fabric Data Agent connected to an e-commerce semantic model and add it as a connected agent to your Sales Associate Assistant, demonstrating how multi-agent architectures can span across Copilot Studio and Microsoft Fabric.

Step-by-step instructions

Create the Fabric Data Agent

  1. Go to fabric.microsoft.com and sign in with your credentials.

  2. In the left navigation, select Workspaces and then select Workshop Demo Workspace.

  3. Within the workspace, locate the e-commerce folder and open the E-commerce Dataset Report to familiarize yourself with the data. Keep this tab open for reference.

  4. Create a personal folder in the workspace by selecting New > Folder and naming it with your username.

  5. In your personal folder, Select New Item, search for data agent, and select Data Agent (Preview).

    Search for Data Agent

  6. Name your agent using the pattern [YourUsername]_DataAgent and select Create.

Connect the Data Source

  1. In the agent setup, select + Data source in the Explorer panel.

    Select Data Source

  2. Select ecommerce-order-dataset (Type: Semantic Model) and select Add.

    Select ecommerce-order-dataset

  3. After validation, select all available tables: Customers, Order_Items, Orders, Payments, Products.

  4. Test baseline functionality in the Test pane:

     What is our total revenue?
 Show me the top 5 products by sales
 How many customers do we have?

Optimize with Meta-Prompts

  1. Generate agent-level instructions by entering this meta-prompt in the Test pane:

     Meta-Prompt: Generate Agent-Level Instructions:
 Analyze your available data sources and create agent-level instructions for yourself (max 15000 chars).

 Objective: E-commerce analytics agent for business intelligence
 Users: Business analysts and sales teams

 Examine your data sources: list all sources, types, and primary use. Analyze domain, time coverage, and main themes.

 Generate instructions with:
 ## Objective
 ## Data Sources (list with priority)
 ## Key Terminology (infer from columns/measures)
 ## Response Guidelines
 Style: Clear summaries with data citations and trend analysis
 ## Handling Common Topics (3-5 based on available data)

  2. Copy the generated response, review it, and paste it into the Agent Instructions field in the Setup section.

    [!TIP] Always review generated instructions for accuracy. Meta-prompts provide a strong foundation but may need minor adjustments for your specific use case.

Publish and Connect to Copilot Studio

  1. Generate a description using this meta-prompt in the Test pane:

     Create a 1-2 sentence description of yourself as a Fabric Data Agent (max 200 chars).
 Analyze your data sources and describe: what data domain you cover and what questions you answer.
 Output plain text only.

  2. Select Publish and paste the generated description in the purpose and capabilities field.

[!IMPORTANT] Save this description — you'll need it when connecting to Copilot Studio.

  1. Go to your Sales Associate Assistant in Copilot Studio.

  2. Select Agents tab on the top navigation bar.

  3. Select Add an agent.

  4. Select Connect to an external agent and choose Microsoft Fabric (preview).

  5. If it shows Not connected, Select the dropdown, select Create new connection, and sign in.

    Create connection

  6. Select your Fabric Data Agent from the list (look for your username) and Select Next.

  7. Configure the connected agent:
    • Name: Your data agent name
    • Description: Paste the description from step 13

  8. Verify the connection shows a green checkmark and Select Add agent.

    [!NOTE] If you see "An unexpected server error occurred", try selecting Add agent again. If that doesn't work, select Back, reselect the agent, and try again.

Test the Integrated Solution

  1. Open the Test pane in Copilot Studio.

  2. Select your market (US or UK) when prompted.

  3. Test the Fabric Data Agent connection with progressive queries:

     What products do we sell in our e-commerce store?

     Show me our top customers by total purchase value

     Analyze our quarterly sales performance and identify which product categories are trending up or down

  4. See in the test pane that data queries are delegated to the Fabric Data Agent, indicated by handoff messages in the execution trace.

    [!TIP] Your Sales Associate Assistant now has three types of agents working together: child agents for market-specific product info, a connected agent for account/contact data, and a Fabric Data Agent for e-commerce analytics. This demonstrates real-world multi-agent orchestration at scale.

🏅 Congratulations! You've completed Use Case #4!

Test your understanding

  • Can you identify when the Fabric Data Agent is being triggered versus the Account Data Lookup Agent in the test pane?
  • How does the agent description influence when the Fabric agent is consulted versus the child agents?
  • What types of queries benefit most from the Fabric Data Agent versus native Copilot Studio capabilities?

🏆 Summary of learnings

True learning comes from doing, questioning, and reflecting—so let's put your skills to the test.

To maximize the impact of multi-agent configurations:

  • Child agents group related capabilities — Use them to logically organize tools and knowledge with dedicated instructions and conditions for orchestration.
  • Connected agents extend your reach — Use them to integrate independently published agents, including agents from other teams or platforms like Fabric.
  • Conditions and variables drive orchestration — Global variables and conditional availability ensure the right agent handles the right request.
  • Descriptions are critical for orchestration — Clear, specific descriptions help the orchestrator decide which agent to delegate to.
  • Test incrementally — Start with simple queries and progressively increase complexity to validate that orchestration routes correctly.

Conclusions and recommendations

Multi-Agent golden rules:

  • Always turn off model knowledge and web search on parent agents when you want strict control over information sources.
  • Use child agents when you need to group tools and knowledge with conditional availability inside a single agent.
  • Use connected agents when you need to integrate an independently managed and published agent.
  • Write clear descriptions for every agent — the orchestrator relies on them to make delegation decisions.
  • Set global variables early in the conversation to establish context that child agents can use for conditions.
  • Test each agent independently before testing the full multi-agent orchestration.

By following these principles, you'll build scalable multi-agent architectures that distribute work effectively across specialized agents, giving users accurate, contextual answers from the right source every time.

📄 Download Options

📄 Download PDF Version Professional PDF with table of contents and internal navigation. Generated automatically from lab content.

📋 On This Page