If I think an LLM is good for something I create well defined, very deterministic "middleware" for that purpose on top of Openrouter.
Anthropic even says, that an agent based solution should only be your last resort and that most problems are well served with a one-shot.
https://www.anthropic.com/engineering/building-effective-age...
I'm much more agreeable with that type of LLM workflow. Running "agents" with monolithic "harness" for long time horizon tasks seems wasteful, unecessary but probably super appealing to lazy people.
Agent use can be used to improve quality and maintainability