NexusAI: Asynchronous AI Placeholders for Minecraft Servers
Server owners are constantly looking for ways to make their worlds feel alive without hiring a team of writers. NexusAI solves that problem by turning AI models into simple placeholders that any menu, scoreboard, hologram or chat plugin can read. It is an asynchronous bridge between your Minecraft server and an OpenAI-compatible endpoint, and it is built for admins and plugin developers rather than for players who want a chatbot to talk to.
What NexusAI Actually Does
The core idea is simple. Instead of asking a model for text every single time a placeholder is read, NexusAI manages answer pools and caches so your server never waits on an HTTP request during gameplay. Two placeholder types do the heavy lifting:
- %ainexus_generate_<prompt>% pulls one stored answer from a unique pool. Each read consumes an answer, and the pool refills in the background, so players rarely see the same line twice.
- %ainexus_cached_<prompt>% uses a shared TTL cache, which is ideal for holograms and scoreboards where the same text should stay stable for a while.
Long prompts can be saved in prompts.yml under a short id, so %ainexus_cached_survival_tips% is far easier to type than a full instruction block. Built-in variables such as {player}, {world}, {biome}, {time} and {weather} work without PlaceholderAPI, and prewarm variables like {player_name} are converted per player.
Format presets keep output clean. You can choose chat, gui, name, hologram, actionbar or bossbar formatting, each with its own length limit and markdown stripping. Reply cleaning removes colour codes, hex colours and MiniMessage tags before another plugin displays them, and the sanitize.allow-markup option stays off by default for good reason.
Supported Versions and Loaders
NexusAI runs on Paper and Purpur from 1.20.6 through 26.2 with Java 21 or newer, and a single jar covers every one of those versions. PlaceholderAPI 2.11.6 or later is a soft dependency and is required for the %ainexus_*% placeholders to resolve. Folia is not supported, and neither are Spigot, CraftBukkit or plain Bukkit. On CurseForge the plugin appears under Bukkit Plugins simply because that is the category for server-side plugins, not because it runs on Bukkit itself.
If you would rather skip manual file management, the foxygame.net launcher keeps NexusAI and your other server add-ons in one catalog, so a fresh build is ready in a single click and version mismatches are handled for you.
Providers, Queues and Reliability
NexusAI speaks to any OpenAI-compatible provider: openai, groq, cerebras, gemini, deepseek, openrouter, ollama or a custom URL. You can register several keys per provider, reference them with ${ENV_VAR}, and rely on masked logging so secrets never appear in console output. Local endpoints such as Ollama need no API key at all.
The model queue is where things get serious. When a provider returns HTTP 429, throws an error, drops below its rate-limit budget or hits a daily cap, NexusAI switches to the next provider or model and returns after the cooldown. There is a fallback model for when the entire queue is unavailable, plus per-server and per-player rate limits and per-prompt backoff.
Dialogues, Actions and Knowledge
The /nai talk <id> command starts a conversation with a character defined in prompts.yml, complete with memory and per-player limits. NPC plugins can trigger the same dialogue from the console. Characters can fire only the commands you explicitly define, with cooldowns, daily limits and permission checks. A knowledge folder of markdown files lets you feed lore and server rules into every prompt.
Optional chat moderation is off by default. When enabled, flagged lines are sent to staff for review, and no player is ever punished automatically.
How to Install NexusAI for Minecraft
Setup takes only a few minutes. Start with Paper or Purpur 1.20.6–26.2 and Java 21, then add PlaceholderAPI 2.11.6 or newer. Drop the jar into your plugins folder and set NEXUSAI_API_KEY or providers.<id>.api-key in the config, using ${ENV_VAR} if you prefer environment variables. Restart the server, or run /nai reload after editing the config. If you use Ollama or another local endpoint, no key is needed.
Before you download NexusAI, remember that prompt text, system prompts, knowledge files and any player values you include are sent over HTTP to the endpoint you configure. With moderation enabled, checked public chat lines are sent too, minus the player name. NexusAI itself collects no telemetry.
Final Thoughts
NexusAI is a focused tool: it feeds AI text into placeholders, dialogues and menus without blocking your main thread. The player-input boundary and reply filter reduce prompt-injection risk but do not remove it, and JSON text components inside replies are not stripped, so keep a stronger model later in the queue and review your prompts carefully. For admins who want dynamic, generated content on a Paper or Purpur server, it is a practical and well-engineered choice.
--- **Update Oct 5, 2026:** Added 1 file for version 26.2, 26.1.2, 26.1.1, 26.1, 1.21.9, 1.21.8, 1.21.7, 1.21.6, 1.21.5, 1.21.4, 1.21.3, 1.21.11, 1.21.10, 1.21.1, 1.21, 1.20.6.