microsoft/agent-framework ? reverse-engineered prompt

I need a framework that lets me build and manage my own AI agents. The core idea is to create individual agents and then connect them together into more complex, multi-agent workflows. It should be like creating a flowchart where one agent can hand off its results to another, or even wait for a human to approve a step before continuing.

I'll be working mostly in Python, but it's important that the framework has full support for .NET as well. A key feature I want is a developer UI where I can test my agents, debug the workflows, and visually trace what's happening step-by-step. It should be flexible, allowing me to plug in different LLMs from providers like Azure or OpenAI. Make sure to include solid observability and logging tools so I can monitor performance and troubleshoot any issues.