To illustrate this, I’d try to do a “demo a day” over the next week to showcase capabilities of the framework. The video below shows an sample app (DevUI) that ships with Agent Framework for running and debugging your Agent Framework tools.

If you are starting a new AI Agent project - try it out!

Ofcourse, frameworks are only there to simplify aspects of your work - what matters is what you build!

- Bring your own model, integrate with any agent
- Workflow API that lets you express business or application logic as a computation graph (this is something that was a weak spot for AutoGen)

Some highlights

- Middleware support - write hooks to control agent behaviour great for RAI, security, context engineering etc
- Support for Agents backed by various services - Azure Foundry Agent, OpenAI Responses, Google A2A agents, ofcourse with full MCP support

Agent Framework brings the production ready focus of Semantic Kernel with full C#/Python support, as well as the nimble intuitive interface of AutoGen for rapid experimentation.
github.com/microsoft/ag...

There is still more work to be done, and we are all excited to make progress with the community!

I will be writing more and showing more examples - afterall a framework is only useful to the extent that it can be used to easily express ideas.

Async stack, event-based messaging, improved observability and control, a low level extremely flexible core api and a high level api with presets, and an improved autogen studio!

With AI, we can have software creators that dont understand complex software systems they create. We’d need to explore co-creation setups that minimize this sort of outcomes.

Yes, drivers dont know how car engines work, pilots probably dont know the details of airplane engines … but mechanics and jet engine designers do.

Once it is complete and communicated well, it looks straightforward and simple.

It is not.

- How to build long-running agents with a complete Python implementation (an MCP server with a travel agent and research agents that stream updates, request input and are resumable.)

The article covers:

- How to build agent-to-agent communication with MCP capabilities where MCP hosts and tools both act as intelligent agents
- Four key capabilities that make MCP tools “agentic” – streaming, resumability, durability, and multi-turn interactions

[1] MCP For Software Engineers | Part 1 : Building Your First Server newsletter.victordibia.com/p/mcp-for-so...
[2] Can You Build Agent2Agent Communication on MCP? Yes! developer.microsoft.com/blog/can-you...

- Resources: Defining server resources, client operations (list/read/subscribe), and real-time update notifications
- Prompts: Creating reusable LLM prompt templates that can be used to modify host application behavior without modifying client/host application code.