M365 Custom Engine Agent

TL;DR: If you already have a multi‑agent AI application, you can surface it inside Microsoft 365 Copilot without adding another orchestration layer. Use a thin “proxy agent” built with the Microsoft 365 Agents SDK to handle Copilot activities and forward a simple request to your existing backend (in this example, we will use a simple Semantic Kernel multi‑agent workflow that writes and SEO‑optimizes blog posts). Develop fast and deploy to Azure with the Microsoft 365 Agents Toolkit for VS Code.

Table of Contents

1. Project Structure
2. Prerequisites
3. Reference Architecture
4. Backend: Semantic Kernel + Azure OpenAI
5. Backend Deployment
   • Docker Container
   • Standalone App
6. Copilot Deployment
7. Licensing
8. Contributing

Project Structure

backend/                Python backend service exposing a Semantic Kernel multi-agent workflow through FastAPI
cea-proxy-ts/           TypeScript based proxy agent to receive messages from Copilot and forward them to the backend
cea-proxy-py/           Python based proxy agent to receive messages from Copilot and forward them to the backend

Prerequisites

1. AI Foundry Project
2. Azure OpenAI deployment (preferably with a model supporting OpenAI Responses API)
3. M365 Copilot License
4. Sideloading is enabled in Microsoft Teams
5. M365 Agents Toolkit Extension

Reference architecture

High‑level flow:

1. Microsoft 365 Copilot Chat → user selects your agent and submits a prompt.
2. Azure Bot Service → delivers Copilot activities (i.e. user messages) securely to your bot endpoint.
3. Proxy Agent (Microsoft 365 Agents SDK) → authenticates, handles activities, and forwards a simple, uniform request to your backend—no extra orchestration here.
4. Your CEA backend → runs your own multi‑agent workflow and returns results for the proxy to relay back to Copilot.

This keeps Copilot integration thin while preserving your orchestration choices.

Sequence diagram:

sequenceDiagram
  participant User as Copilot User
  participant Copilot as M365 Copilot Chat
  participant ABS as Azure Bot Service
  participant Proxy as Proxy Agent (Agents SDK)
  participant Backend as CEA Backend (Semantic Kernel)

  User->>Copilot: "Write a blog post about X, optimized for SEO"
  Copilot->>ABS: Activity (message)
  ABS->>Proxy: POST /api/messages (JWT)
  Proxy->>Backend: POST /generate-blog { prompt }
  Backend-->>Proxy: { content }
  Proxy-->>ABS: sendActivity(content)
  ABS-->>Copilot: Reply
  Copilot-->>User: Render AI response

Loading

The Backend with Semantic Kernel Agents + Azure OpenAI

The backend (backend/main.py) is a FastAPI app using Semantic Kernel’s Azure Responses Agents with Azure OpenAI. It exposes one uniform endpoint consumed by the proxy:

• POST /generate-blog
  • Accepts: { topic, length } or { prompt } (free‑form)
  • Returns: { content: string } (final markdown)

Sequence diagram of the internal multi‑agent workflow:

sequenceDiagram
  participant Param as ParamExtractionAgent
  participant Writer as BlogWriterAgent
  participant SEO as SEOOptimizerAgent

  Param->>Writer: topic, length
  Writer->>Writer: Generate draft
  Writer->>SEO: Draft article
  SEO-->>Writer: SEO guidance (structured)
  Writer-->>Writer: Finalize with SEO
  Note over Writer: Final markdown content

Loading

Multi‑agent flow:

1. Parameter Extraction Agent

   • Input: Free‑form prompt
   • Output: JSON with topic and length (clamped; default 5)
   • Model: Small, low‑temperature for deterministic structure

2. Blog Writer Agent

   • Input: topic, length
   • Output: Markdown draft
   • Model: Creative settings for richer content

3. SEO Optimizer Agent

   • Input: The draft
   • Output: Structured JSON (title, meta, slug, headings, revised article, improvements, keywords, links, readability, CTA)
   • The structured SEO output is fed back into the writer for a final pass

Backend Deployment

Prerequisites

1. Rename sample.env to .env
2. Fill in the required values:

AZURE_OPENAI_ENDPOINT=
AZURE_OPENAI_API_KEY=
AZURE_OPENAI_CHAT_DEPLOYMENT_NAME=
AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME=
AZURE_OPENAI_API_VERSION=

[Option 1] Run backend in a container

⚠️ Requires Docker Desktop or a similar Docker engine.

From repo root:

cd backend
docker build -t backend-blog-agent:latest .
docker run -d --env-file .env -p 8000:8000 --name backend-blog-agent backend-blog-agent

[Option 2] Run backend as a standalone app

⚠️ Make sure you are using the uv package manager

From repo root:

cd backend
uv venv
uv sync
uvicorn main:app --host 127.0.0.1 --port 8000

Copilot Deployment

For demonstration purposes, the proxy agent in this sample is provided in two different flavors: Python and TypeScript. The process to run your Custom Engine Agent is exactly the same.

1. Open cea-proxy-ts or cea-proxy-py as the root folder in Visual Studio Code

2. Open the M365 Agents Toolkit

3. Sign-in to M365 and Azure

   

4. Select Debug with your local Environment. This will:

   • Create a devtunnel to emulate an Azure Bot and forward requests to your CEA
   • Register a App registration in Entra ID which is tied to your Bot service
   • Register the newly created Bot in Teams and Copilot
   • Open a browser window to Copilot and launch your Copilot Agent

5. Enter your prompt, provide a blog post topic and indicate its length.

   

Licensing

MIT License (see LICENSE).

Contributing

See CONTRIBUTING for details.