I just changed this sample python app to grafting reducing integration layer code by 67%, decoupling code from integration method and making backend API a pure public methods which now can be called through any channel and immediately support MCP as well.

It gives different benefits: less code > easier to understand and maintain by AI > much less tokens used (compared same prompt on grafting and without - what is obvious without even checking as adjusting or understanding public method vs public method + fast api wrapper + client is obvious win) > easier to review pull requests from AI > implementation match simple OOP UML class, interaction and sequence diagrams > ability to change architecture of integration via env var in production even changing from http2 to WebSocket to azure ServiceBus queue etc…

All is based on our solution that now is achieving stable release with full open telemetry support but already gives all you need for majority of projects.

Same will work for you across netcore, Java, kotlin, groovy, python, ruby, php, node, js/ts in browser. Ruby and php is in more alpha state.

If you want to try create empty folder open in cursor run:

Windows: iwr grft.dev/get | iex

Linux/mac: curl -fsSL grft.dev/get | sh

And choose ai rules > cursor and next just type your prompt to get any distributed system you will see magic of grafting on your own sample :)

Let me know your thoughts!

Bests, Przemek graftcode co-founder and CEO