FAQ

Before you get the zip, the converted server is started headless in an isolated Docker sandbox and has to reach Minecraft's own Done (…s)! ready line. If it crashes, the engine reads the log, applies a deterministic fix (remove the client mod that slipped through, add a missing library, adjust memory), and boots again - up to five attempts by default. Only a pack that reached "Done" gets a download button.

In order of trust: what the pack itself declares (.mrpack env metadata) → what Modrinth says about the project's server side → what the mod jar declares about itself → a curated list for the known troublemakers. Every removal lands in the report with its reason and its source - and you can override any of it with the "always keep" and "also remove" boxes in the converter's options.

You get the boot log, what the engine tried, and a diagnosis when a rule matched. Some packs genuinely need hand-tuning. Bring the log to the Discord - failed conversions are how the rules database grows.

Forge, NeoForge and Fabric, roughly Minecraft 1.16.5 through current. Quilt is on the list. The boot sandbox picks Java 8 / 17 / 21 / 25 automatically per Minecraft version.

Mods are fetched through the official Modrinth and CurseForge APIs at conversion time. When a CurseForge author has opted out of third-party downloads, Packworks doesn't fetch the file at all - it pauses so you can download it from the author's own page (their download counter and rewards stay intact), verifies your file by size and hash, and continues. Skipped mods ship as a small fetch-script instead of a jar.

Never. Mods are their authors' work and always come from their official channels. What Packworks sells (later - see pricing) is the assembly, configuration, testing and packaging labor.

Nothing. No account either. After launch there are exactly two options: $3.99 a month or $39.99 once for lifetime access, both with every feature. People on the notify list hear first.

One click in account settings, no chat with a retention bot. That's a design decision, not a promise of convenience.

It lives on the server only as long as the job needs it, then gets cleaned up by a scheduled prune (about two weeks, so you can re-download the artifact). Logs pasted into the Doctor are analyzed in memory and not stored.

One machine, a deterministic rules engine, and Docker. No AI is involved in conversions - results are reproducible, auditable, and cost nothing per job, which is why the beta can be free.