Why this exists

A friend of mine runs a small Rust server. Ten players at the most, mostly her mates. By the time she told me about it she had spent about $3,500 on mods for it. Half of them did not work. The ones that did fought each other, the server ran like a wet sock, and it was honestly unplayable when I joined. She is not wealthy, and there was never going to be a return on that money. It was just gone.

That was the start. The first thing she wanted to buy after that was another mod of the same kind. I said, why not make one instead? That became Salvo, the first MixMods mod. Then the next thing she needed, then the next. Many months later that pile of one-at-a-time fixes had become a server pack that runs on one framework, shares one interface, and does not fight itself. It runs on our two servers today, through every weekly wipe, with real players. It is still a work in progress and I expect it always will be.

I have about ten thousand hours in Rust. I like the game. What I do not like is that running a small server for your friends costs more in mods than it costs in hardware, and that the free route means assembling dozens of plugins yourself and finding out on a live wipe which ones clash.

How the mods get made

I use AI to write code. I am not going to pretend otherwise, and I am not going to pretend it is easy. There is no twenty-minute prompt that produces a mod. Anyone who tells you there is has not shipped one.

What the tool does is let one person finish the whole brief. I know the game and what a server needs; I do not have an engineering team. So I oversee, I test, I play it, and the parts I cannot do alone get done with help, then torn apart until they hold. The first results were not good enough, and a lot of the last year went into building the testing that catches what I could not see. That system is the real product. The mods are what comes out of it.

How I test

I am not publishing the tooling yet, but I will say exactly what it checks, because that is the difference between a method and guesswork.

• Every file is compiled two ways against the current Rust build before it goes anywhere: on the Carbon framework, and again without any compatibility shims so it is not leaning on a crutch that the next game update removes. The result is public on every listing here that hosts a plugin file, dated and stamped with the game build.
• Nothing deploys to our servers without passing a gate that fails on cost: hooks that fire too often, work done on the hot path, timers under a second, walks over every entity, wrong hook signatures. Style is a warning. Performance is a fail.
• Load is measured, not guessed. We drive the pack with phantom players inside the real server, walking, fighting, looting, chatting, and read the frame time. Measured 2026-09-26 on our Australian box, Rust build 25083359: with 200 phantom players the full pack used under one percent of frame time.
• Every plugin gets independent cold reads before release, by a reviewer who has not seen it before, against a written checklist. The 2026-09-19 pack release took eighteen rounds of that.
• It runs on our own servers before it goes in the pack, with a probe that presses every panel and command from a player's seat.

None of that makes a mod "safe" or "verified". It makes it checked, and it makes the checking visible.

What it costs

Two top-tier AI subscriptions and two servers cost me about $1,000 AUD a month. That is the whole bill, and it is a lot for one person.

Free. It is in beta: feedback and testers wanted. No feature is behind a payment and no fix ever will be. If the work is useful to you and you can spare something, there is a support link. It goes on the bill above and nothing else. If the bill is not covered, the servers still run and the mods still work; what stops is new development.

That is the deal, and it is the one I wish someone had offered my friend.