Project: [Enterprise Agents] - [Whatever AI Assistant (WAIA)] #52
Description
Track
Enterprise Agents (M365 Agents Toolkit)
Project Name
Whatever AI Assistant (WAIA)
GitHub Username
Repository URL
https://github.com/FranckyC/agents-league-whatever-ai-assistant
Project Description
A production-ready, extensible multi-agent system for Microsoft 365 Copilot Chat and Microsoft Teams. It showcases a flexible pattern for common cross-domain company agents (IT and HR here as an example), covering both question-and-answer scenarios (RAG pattern) and action automation (e.g., IT ticket submission). Built with Microsoft Foundry Agent Service, Microsoft 365 Agents SDK, and a custom MCP server.
Key Highlights
| Feature | Description |
|---|---|
| Multi-Agent Routing | A workflow agent routes queries to specialized HR, IT, or Fallback agents — transparently to the user |
| Streaming Responses | OpenAI Responses API + Teams streaming for a smooth conversational experience |
| MCP Server with OAuth | Custom MCP server secured via OAuth Identity Passthrough — delegated permissions only |
| Read & Write MCP Tools | copilot_retrieval for SharePoint knowledge retrieval via the Copilot Retrieval API · submit_ticket for IT ticket submission in a SharePoint list |
| Adaptive Cards | Debug card, ticket form, disclaimer, and MCP tool approval — all as rich interactive cards |
| Custom Disclaimer | Dynamic Adaptive Card disclaimer on first interaction or new chat sessions |
| Human in the Loop | MCP tool approval flow via dynamic Adaptive Cards |
| Built-in Debug Mode | /debug on|off command to inspect agent reasoning directly in Teams/Copilot |
| OpenTelemetry | Telemetry integration for MCP tool calls, responses, and errors — visible in the Foundry portal |
| Citations | Proper handling of agent references for both Teams and Copilot experiences |
Demo Video or Screenshots
Demo video: https://youtu.be/LGkXO8tr1Fg
Screenshots: https://whatever-ai-assistant-docs.azurewebsites.net/
Primary Programming Language
TypeScript/JavaScript
Key Technologies Used
- Microsoft Foundry Agent Service with workflow agent
- Model Context Protocol with OAuth Identity Passthrough
- Microsoft 365 Copilot
- Microsoft Graph (Copilot Retrieval API and SharePoint)
- TypeScript/Node.js
- Microsoft 365 Agents SDK and CLI
- Bicep templates and Azure deployment stack
- Docusaurus :D
Submission Type
Individual
Team Members
No response
Submission Requirements
- My project meets the track-specific challenge requirements
- My repository includes a comprehensive README.md with setup instructions
- My code does not contain hardcoded API keys or secrets
- I have included demo materials (video or screenshots)
- My project is my own work with proper attribution for any third-party code
- I agree to the Code of Conduct
- I have read and agree to the Disclaimer
- My submission does NOT contain any confidential, proprietary, or sensitive information
- I confirm I have the rights to submit this content and grant the necessary licenses
Quick Setup Summary
Steps (high level)
- Install prerequisites
- Prepare the data
- Deploy infrastructure in Azure
- Configure MCP in Foundry
- Set up agents in Foundry
- Package and upload the Teams app
- Talk to the agent
Full getting started procedure available here: https://whatever-ai-assistant-docs.azurewebsites.net/getting-started/
Technical Highlights
- Leveraging custom MCP server (along the bot logic, reusing the same M365 Agents Toolkit development tooling) with Foundry Agent Service OAuth Identity Passthrough to call Microsoft Graph API using user identity (delegated).
- Flexible architecture
- Agent flow integration with Adaptive Cards for authentication and MCP tool approval prompts.
- Usage of custom prompt marker to send forms from Teams/Copilot channel and pass the data back to agent.
- Debug mode to inspect agent output right from the Teams or Copilot channel.
- Replicable end-to-end automated deployment to use the solution in any Azure environement.
Challenges & Learnings
- Integrate agent LLM flow with Teams/Copilot experience (adaptive cards, dialog loops, etc.).
- MCP tools/Agent prompts crafting to get a reliable path.
More info about challenges and strategies here: https://whatever-ai-assistant-docs.azurewebsites.net/dev-notes
Contact Information
Country/Region
Canada