Hello Merlin Team, I greatly appreciate your app for providing multimodal for a reasonable price. Couple of features I think would be a great addition for future.

-- A windows desktop app with features that can help build apps using these models.

-- A CLI that can work anywhere on computer.

-- Noticed the app's response is pretty fast using pro models which is great compared to actually using the same model on Claude for examples using Opus. Not sure if app is set to use fast mode that consumes credits faster. I see the credits finish as soon as a single query is completed. Either changing the mode or letting user aware that how many tokens might be consumed based on the input would help a lot to reduce context or use efficiently.