I wanted to share two open source .NET repos I’ve been building around OpenAI agents and ChatKit-style workflows.
The gap for C# developers has been pretty obvious. Most of the examples and working code in this area show up first in Python or JavaScript. So from the .NET side, you usually end up translating the pattern yourself, then doing extra work to make it fit a normal app, DI setup, hosting model, and so on.
So I split the work into two repos.
incursa/openai-agents-dotnet
This is the agent side. It is my .NET implementation around the Agents SDK-style concepts, built to feel like normal C# instead of a straight port.
incursa/chatkit-dotnet
This is the ChatKit side. It covers the ChatKit-style usage and hosting layer in .NET, and it sits next to the agents repo instead of mixing those concerns together.
These are not official OpenAI libraries. They are community .NET projects.
I built them because I wanted something I’d actually want to use in a real codebase: clear boundaries, APIs that make sense in C#, and a shape that fits normal .NET applications without dragging in a bunch of assumptions from another ecosystem.
They’re still evolving, but that is part of the point. I’d rather get the structure right now than rush to cover everything.
The feedback I’d most like is pretty practical: does the split between the repos make sense, where does the API shape feel off for .NET, what is still missing for real application use, and where should a .NET implementation stay close to upstream versus take the more natural C# path?
If you’ve worked with the Python side, the JavaScript side, or you’ve been trying to do this from .NET, I’d like to hear what looks right and what doesn’t.