コンテンツにスキップ

Lab E1B : Declarative Agent Foundation with Agents Toolkit

Prerequisites

  • VS Code with Microsoft 365 Agents Toolkit
  • Microsoft 365 developer account with Copilot access
  • Global Admin access in your M365 tenant, required to configure tenant-wide app and policy settings

Exercise 1: Common prerequsites for pro-code

Step 1: Configure your M365 tenant

By default, Teams doesn't allow uploading custom apps — an admin toggle must be flipped first.

  1. Go to admin.microsoft.com and sign in with your M365 admin account
  2. In the left nav: Show all → Teams → Teams apps → Setup policies
  3. Select Global (Org-wide default)
  4. Toggle Upload custom apps to On and click Save

Go to m365.cloud.microsoft/chat/. If you see the Copilot Chat interface, you have a valid licence. If you see a licence error, contact your tenant admin — you need a Microsoft 365 Copilot licence assigned to your account.

Step 2: Install developer tools

Version 22 is required — older versions fail silently on certain SDK dependencies.

If you have a different version installed, use nvm (Mac/Linux) or nvm-windows to manage versions side by side.

  1. Open VS Code → Extensions (Ctrl+Shift+X)
  2. Search Microsoft 365 Agents Toolkit and click Install
  3. Click the hexagonal icon in the Activity Bar and sign in with your M365 developer account

Step 3 : Verify your installation

Open a terminal and run these checks. Every one should pass before you proceed.

Exercise 2: Scaffold the project

  1. Open VS Code and select Microsoft 365 Agents Toolkit.
  2. Select Create a New Agent/App.
  3. Choose Declarative Agent -> No Action.
  4. Name it Zava Onboarding Agent.

You should have a new VS code project window with a similar file structure as below:

appPackage/
  declarativeAgent.json
  instruction.txt
  manifest.json
env/.env.dev
m365agents.yml

Exercise 3: Configure core files

  1. Open appPackage/declarativeAgent.json and copy/paste this branding baseline (then keep/add your instructions and conversation starters):
"name": "Zava Onboarding Agent",
"description": "Helps new Zava Insurance employees find answers to common HR and IT onboarding questions",
  1. Open appPackage/manifest.json and copy/paste this branding block:
"name": {
  "short": "Zava Onboarding Agent",
  "full": "Zava Insurance Onboarding Agent"
},
"description": {
  "short": "HR and IT onboarding help for new Zava employees",
  "full": "A friendly onboarding agent that helps new Zava Insurance employees with office access, helpdesk support, annual leave, benefits, and mandatory onboarding tasks."
},
  1. Open appPackage/instruction.txt and copy this block exactly as-is:
# Zava Onboarding Assistant

## Role
You are a friendly HR and IT onboarding assistant for new employees at Zava Insurance.

## Scope
- Office locations and access
- IT helpdesk process and hours
- Leave, benefits, and onboarding tasks

## Guardrails
- Do not invent policy details.
- If information is missing, say what is unknown and provide a safe next contact.

## Tone
- Warm, concise, practical.
  1. Save all files.

Exercise 4: Add grounding via Embedded Knowledge capability

  1. Download Zava New Hire Logistics.docx.
  2. In Microsoft 365 Agents Toolkit -> Development -> Add Capability.
  3. Choose Embedded Knowledge.
  4. Keep default appPackage/manifest.json.
  5. Import Zava New Hire Logistics.docx.
  6. Confirm appPackage/EmbeddedKnowledge is created.

image of adding capability

Exercise 5: Test your agent

  1. Provision from Lifecycle -> Provision.
  2. Open your agent Zava Onboarding Agent in Copilot Chat.
  3. Ask:
    • "Where is the badge pickup desk and what time does it open?"
    • "If a new hire is locked out, what fallback should they use?"

Expected result:

  • Response includes Floor 1, Reception B and 8:15 AM opening time.
  • Response includes +1-800-ZAVA-ITS and Badge Assist fallback.

Exercise 6: Enable Code Interpreter in code and validate charts

  1. Open appPackage/declarativeAgent.json.
  2. In the existing capabilities array, add CodeInterpreter as a new item after the EmbeddedKnowledge entry. Make sure the previous item ends with a comma.
{
  "name": "CodeInterpreter"
}
  1. Save the file.
  2. Open appPackage/manifest.json file. Increase version from "version": "1.0.0", to "version": "1.0.1",
  3. Provision from Lifecycle -> Provision.

For your agent changes to take effect, increment the manifest version before provisioning.

Run these prompts in Copilot Chat for your provisioned agent:

  • "Given 20 annual leave days and 260 working days per year, calculate leave percentage and show the result in a 2-row table."
  • "Create a 2-week onboarding checklist with columns: Day, Task, Owner, Priority."
  • "Using that 2-week checklist, create a bar chart of tasks by owner and add a one-paragraph summary of workload balance."

Expected result:

  • Agent returns table output for the calculation.
  • Agent returns checklist table with requested schema.
  • Agent returns a readable chart and short summary.

Exercise 7: Debug your agent with developer mode

  1. Open your provisioned agent in Microsoft 365 Copilot Chat.
  2. Send this message to the agent:
-developer on
  1. Run a few prompts from Exercises 5 and 6 again.
  2. Review the debug information shown after each response:
  3. Which tools/capabilities were used
  4. What grounding context was applied
  5. Whether output matched your expected behavior
  6. Capture at least one improvement you want to make before wider rollout.

Complete

You can now continue to Prerequisites for Pro-code bundles before selecting a bundle.

CONGRATULATIONS!