U-he preset randomizer / merger (open-source CLI tool)

[Fannon]

Claude should do a great job understanding semantic relationships from metadata.
I wonder how close you could get to the same functionality using a Skill vs. MCP?

Yes, I also think LLMs tend to be good at this, but sometimes they are overconfident and overlook things. For example in one case I checked it totally overlooked how strongly the cutoff would affect the sound.

Under the hood I use something like skills that are specific to each u-he synths, which bakes in some knowledge how their presets and parameters work (generated, but based on statistics from the analyzer). So this MCP server has sub-skills included so to say.

I'm not sure this can be a skill, because it's not just about adding context to an LLM, there is actually quite some deterministic code involved that accesses your file system, finds your synths, reads presets into a library, does some statistical distribution analytics and then can randomize / breed presets from that. But I totally get the question! I think with Claude skills there are now quite a few things that should be just a skill!

[AtomOfScent]

Under the hood I use something like skills that are specific to each u-he synths, which bakes in some knowledge how their presets and parameters work (generated, but based on statistics from the analyzer). So this MCP server has sub-skills included so to say.

I'm not sure this can be a skill, because it's not just about adding context to an LLM, there is actually quite some deterministic code involved that accesses your file system, finds your synths, reads presets into a library, does some statistical distribution analytics and then can randomize / breed presets from that. But I totally get the question! I think with Claude skills there are now quite a few things that should be just a skill!

Oh, interesting.

I didn't know the MCP server capability was an update to the original repo.
I'll be sure to have a closer look.

[Fannon]

If you try it out, let me know what you think of it. I'm not sure if this is really useful / sensible, but it was a fun experiment to try.

[marcoskohler]

I tried installing this today and using it, with the help of GPT, and no luck. I submitted an issue on the github repo