SfxKit

Sounds are a big deal. Not so much in web-dev, because it isn’t the early 2000’s where every button needs to bleep. In games though? Feedback is king. And the earlier you consider sounds, the better. I’ve had to rewrite entire sections of code to accommodate sounds because I didn’t think about it during the initial implementation. There’s a few websites that have thousands of open source clips. Finding the right clip this way might be the worst part of game development for me, so I avoid them like the plague. (I’m looking at you Freesound.org) So. With no other choice, I was generating SFX from a little website with presets and sliders that was good enough for the retro style of my game. It wasn’t frictionless though. So eventually, I had an idea.

Approach

I discovered the synth I was using was actually Open Source, and pretty simple. So all that was left was to pull it down, port it to C#, wrap it in a CLI and a Skill, and I could just tell my agent to make sounds for my Unity project. I was so impressed during early testing, I decided to pull it out into a Unity-free module and create a dedicated MCP pipeline for the skill to handle my main case. Now I can have retro sounds whenever I want.

The MCP side of things is nice too, I mention the prefab I’m working on, a quick blurb about what I’m expecting, and the agent drives the tool. Once it has the .wav file (or a few), the MCP step kicks in, and my Agent adds the .wav to the project, wraps it in a tiny Scriptable Object to allow for things like random pitch and different ways to play sets of sounds, and I can pick it from the list just a minute or so after I ask.

I was cackling like a madman the whole time.

Hear it

Four sounds generated from a JSON spec. Same input, same output every time.

These were generated with zero human input beyond “I want some sounds to show off on this devlog page.” No tweaking, no auditioning. Spec in, sounds out.