VitaminMCP

Nothing breaks against 3.0.1, and there is nothing to change on upgrade. This release is three fixes, one of which decided whether bots worked at all on a large class of servers.

Fixed

Bots could not join a server whose plugins send a resource pack. The bot authenticated, went quiet, and thirty seconds later bot_spawn failed with did not join within 30000ms — every time, whether it connected through a proxy or straight to the backend. Servers that host their own pack were therefore untestable. Bots now accept every pack the server sends and download none of them, which is enough to be let in. Nothing about a pack is visible to a bot, so checking what is inside one still needs a real client.

A join that does fail now says how far it got, rather than repeating the same sentence for every cause:

Bot Tester1 did not join within 30000ms (last state: configuration, last packet received: keep_alive)

runnerJar pointing at a .jar failed with CreateProcess error=193, an operating-system message that named neither the problem nor the fix. The setting kept its name after the runner stopped being a jar, so paths from older notes outlive the documents that produced them. It is now refused with the names it actually takes.

On Linux and macOS, a runner placed next to mcp-server.jar was not found: session_start without runnerJar recognised only the Windows executable, and reported no runner while it sat in that directory. All five published runners are now recognised. Installs through npx vitaminmcp were never affected.