Conclusion

Congratulations — you have completed the Microsoft Agent Framework Workshop!

What You Built

Starting from a blank Python project, you built OpsAgent — a fully featured AI-powered operations and engineering assistant:

Module	What You Built	Skills Gained
1	Python project, venv, dependencies	Project setup with `uv`
2	LLM client via GitHub Models	OpenAI-compatible client
3	First OpsAgent with instructions	`Agent`, `OpenAIChatCompletionClient`
4	Azure health check & deployment tools	`@tool` decorator, tool calling
5	Microsoft Learn MCP server	`MCPStreamableHTTPTool`, live docs
6	Conversation history & memory	`InMemoryHistoryProvider`
7	Persistent agent memory	File/DB-backed persistence
8	Multi-agent orchestration	`Workflow`, agent pipelines
9	Streamlit, Chainlit, FastAPI UIs	Chat interface patterns
10	Azure Functions HTTP endpoint	`AgentFunctionApp`, durable state

OpsAgent — Final Architecture

                        ┌─────────────────────────────────┐
                        │           OpsAgent               │
                        │                                  │
                        │  ┌──────────┐  ┌─────────────┐  │
  HTTP / Chat UI ──────►│  │ Tools    │  │  MCP Server │  │
                        │  │ (M4)     │  │  (M5)       │  │
                        │  └──────────┘  └─────────────┘  │
                        │                                  │
                        │  ┌──────────┐  ┌─────────────┐  │
                        │  │ History  │  │  Workflows  │  │
                        │  │ (M6/M7)  │  │  (M8)       │  │
                        │  └──────────┘  └─────────────┘  │
                        │                                  │
                        │     GitHub Models (GPT-4o mini)  │
                        └──────────────┬──────────────────┘
                                       │
                              Azure Functions
                           (Durable HTTP endpoint)

Key Concepts Recap

Concept	Class / Feature
Agent definition	`Agent(client, name, instructions, tools)`
Tool registration	`@tool(approval_mode="never_require")`
MCP integration	`MCPStreamableHTTPTool(url=...)`
Conversation history	`InMemoryHistoryProvider(load_messages=True)`
Serverless hosting	`AgentFunctionApp(agents=[...])`
Multi-turn threads	`x-ms-thread-id` header

Where to Go Next

Microsoft Learn — Agent Framework — Full reference documentation, API guides, and tutorials. learn.microsoft.com/agent-framework
Official Samples — More hosting options, multi-agent patterns, and advanced scenarios. github.com/microsoft/agent-framework
MCP Servers — Browse and connect additional MCP servers to extend OpsAgent. modelcontextprotocol.io
Deploy to Azure — Move from func start to a real Azure subscription with one command. Deploy to Azure Functions

Extend OpsAgent

Some ideas to keep experimenting:

Add a new @tool for querying Azure Monitor logs
Connect a second MCP server (e.g. GitHub MCP for issue triage)
Replace InMemoryHistoryProvider with a Cosmos DB-backed provider
Expose OpsAgent via the A2A protocol so other agents can call it
Add an OpenAI-compatible endpoint so any OpenAI client can use it
Wire OpsAgent into a multi-agent Workflow alongside a CodeReviewAgent

Workshop Complete — You now have a solid foundation for building, extending, and deploying AI agents with Microsoft Agent Framework. The patterns learned here — tools, MCP, history, workflows, and hosting — apply directly to production systems.