I use Claude to make my prototypes within Unreal Engine. I don’t have anything embedded in the engine itself, I have the program installed and go back and forth to implement what the LLM suggests me.
It works pretty good, I have to say. On one end, it forces me to structure well the problem to solve; Claude helps with everything, from research to properly framing. I am faster in creating my own JIRA tasks and message my colleagues in a proper and clear way. Then, I use blueprint and I have to say that the results (from a prototype perspective) are pretty satisfying. Also, at the end of each chat I tell it to create a report for its knowledge base, so that in future should be better and better. At least, I hope so.
There is a challenge though: switching back and forth from Claude to the engine to implement what the LLM says is tiring and I end up my days with headache. It’s like having someone giving you suggestions constantly. I know there is a way to integrate everything into the engine, but it’s still experimental and I don’t want to risk on a shared code.