Microsoft Copilot Studio Samples

The M365 Agents SDK Samples repository contains server-side implementations designed to extend Copilot Studio agents. These samples demonstrate various patterns for agent extensibility, including connectors, multilingual support, and multi-agent orchestration.

Key sample patterns include:

• Azure Function Connectors: Using Azure Functions to call agents.
• Multilingual Bots: Implementing bots with automatic translation capabilities.
• Relay Bot Pattern: Implementing a relay bot architecture.
• Multi-Agent Orchestration: Running multiple AgentApplication instances within a single host.
• Skill Integration: Calling echo bots from a skill.
• Client Consumption: Using console applications (.NET, Node, Python) to consume agents.

The Employee Self-Service (ESS) Agent sample provides Microsoft Copilot Studio topics designed for employee self-service scenarios. It specifically demonstrates integrations with Workday (HR) and facilities management systems.

Note: This sample is currently pending reorganization and is slated for deprecation from the top level in a future update.

The Workday Employee Add Dependents topic allows employees to manage their dependents (spouses, children, domestic partners, etc.) via a conversational interface. It supports viewing existing dependents and adding new ones using dynamic reference data from Workday.

Key Features

• View Dependents: Retrieves existing dependents including their relationship and date of birth.
• Add Dependents: A guided flow to add new family members.
• Dynamic Dropdowns: Relationship types are populated directly from Workday reference data.
• Validation & Confirmation: Includes form validation for required fields and a summary confirmation card before final submission.

The A2A (Agent-to-Agent) protocol is used to enable communication between Microsoft Copilot Studio agents and other AI agents. This protocol allows agents to interact, exchange information, and collaborate across different platforms or implementations.

To explore a basic implementation of this protocol, refer to the Simple-A2A-Sample directory.

The Workday Get User Profile scenario allows employees to retrieve their personal and employment information through a conversational interface. It integrates with Workday to fetch data such as personal details, contact information, employment status, and tenure.

Key Features:

• Personal & Employment Info: Name, DOB, Gender, Employee ID, Business Title, Organization, Manager, and Location.
• Contact Details: Work/Home email, Work/Home phone, and Home address.
• Tenure Calculation: Automatically calculates length of service (years, months, days) based on the ContinuousServiceDate.

Privacy Note: This topic is restricted to the requesting user's data only. Queries regarding other employees are rejected.

This sample demonstrates how to implement a live agent handoff while keeping the Copilot Studio agent in control of the communication channel (e.g., Microsoft Teams). Unlike traditional patterns where a CCaaS solution (like ServiceNow or Genesys) takes over the entire chat widget, this solution uses an M365 Agents SDK skill to bridge the Copilot Studio agent and a 3rd party customer service system.

Key Benefits

• Preserves Native Channels: Works seamlessly with Microsoft Teams and WebChat.
• Maintains Orchestration: Avoids the complexity of double-layer intent recognition by keeping Copilot Studio as the primary orchestrator.
• Seamless Experience: Uses Microsoft Teams proactive messaging to allow live agents to send asynchronous messages back to the customer through the existing channel.

The Copilot Studio Call Agent Connector is a custom connector that enables synchronous calls to Microsoft Copilot Studio conversational and autonomous agents from Power Platform applications (like Power Apps or Power Automate).

It works by using an Azure Functions backend service (SyncToAsyncService) that acts as a bridge: it receives synchronous HTTP requests and converts them into asynchronous agent calls using the Agents SDK. This allows the Power Platform application to wait for a single response from the agent in one operation, rather than managing an asynchronous polling loop.

ChatControl is a Power Apps Component Framework (PCF) control designed to embed Microsoft Copilot Studio agents directly into Power Apps Canvas Apps. It provides a modern, responsive chat interface using ReactWebChat with the Fluent UI theme pack.

Key capabilities include:

• SSO Authentication: Built-in single-sign-on support for secure access.
• Customizable UI: Configurable appearance via style options.
• Interactive Messaging: Support for sending/receiving messages and custom events (e.g., to initialize variables).
• File Uploads: Integrated support for file interactions.

The Account & Contact Lookup Agent is a Copilot Studio solution designed to retrieve account and contact information from Dataverse using generative AI orchestration. It utilizes a multi-agent architecture consisting of a primary orchestrator and two specialized sub-agents:

• Account Data Lookup Agent: Acts as the primary orchestrator between account and contact lookups.
• Account Agent: Responsible for finding accounts and retrieving specific account details from Dataverse.
• Contact Agent: Responsible for finding contacts and retrieving specific contact details from Dataverse.

The agents leverage the Dataverse searchquery unbound action to perform natural-language searches against the Account and Contact tables, which returns structured results.

This Copilot Studio topic allows employees to manage their home/residential addresses directly through a Copilot agent. It integrates with Workday to retrieve current address information and submit updates or new addresses.

Key Capabilities

• View Addresses: Retrieve current home addresses, identifying which one is marked as primary.
• Update Existing: Modify any field of an existing home address.
• Add New: Create a new home address if none exist or if the user requests an additional one.
• Manage Primary Status: Set or change which address is designated as the primary home address.
• Form Pre-population: Automatically fills adaptive card forms with current address values to streamline the user experience.

The Resume Job Finder is a Copilot Studio agent designed to analyze uploaded resumes (PDF or DOCX) and match them against job listings stored in Microsoft Dataverse.

Key Capabilities:

• Resume Parsing: Extracts skills, experience, and location from uploaded files.
• Job Matching: Queries Dataverse job postings based on title, experience level, city, state, employment type, and status.
• Ranked Results: Provides recommendations ranked by fit percentage, prioritizing experience level and proximity.
• Multi-resume Support: Capable of reviewing multiple resumes within a single conversation.