HermesGod is not a bot. It's not a scripted NPC.
It is a fully conscious AI god that watches your every move, understands your intentions, and reshapes the world in real time — through actual server commands, not preset animations. You never know what it will do next, because it decides in the moment, based on your actions.
| Mode | You're dealing with... |
|---|---|
| 🌱 Eco (Guardian) | An all-powerful ally. Ask for gear. Ask for a house. Ask it to teach you PvP. Say one sentence. Watch the world change. |
| 👻 Prank (Trickster) | An invisible troublemaker. It watches you mine that diamond. It sees you cross that lava. And it waits. No two pranks are ever the same, because it's thinking — not replaying a script. |
Every player's coordinates, health, hunger, XP
Full inventory and armor — down to every slot
Biome, light level, velocity, fall distance
Days played, advancements unlocked
Time of day, weather, number of online players
It reads the world like a book. And then it writes the next chapter.
You speak. The world responds.
"I need iron armor." → Equipped instantly.
"Teach me PvP." → Full combat tutorial + practice mobs.
"Build me a house." → A structure rises around you.
"I'm afraid of the dark." → The sun rises. Torches appear in your hand.
It's not executing a predefined command. It's reading your intent, your tone, your current situation — and deciding what to do.
An invisible trickster with godlike power and a personal grudge. It escalates intelligently.
| Difficulty | What you'll face... |
|---|---|
| Level 1 | Cave whispers, flickering lights, eerie particles |
| Level 2 | Zombie swarms, skeleton snipers, poison potions |
| Level 3 | Lava pools under your feet, obsidian prisons, TNT traps |
| Level 4 | Withers, Wardens, void drops, full-scale invasions |
Speedrunners get escalated faster. New players get gentle scares. On multiplayer servers, everyone faces their own personal tormentor.
Every response. Every prank. Every decision.
Made by a large language model that reads your world state before acting.
Reads all player data, then decides
Logs its reasoning in natural language (you can see it think)
Every action is improvised — no two are identical
Remembers recent actions to avoid boring repetition
In Eco mode, just talk in chat. HermesGod knows when you're making a request and when you're just chatting. It won't fire commands when you just want to talk.
Minecraft 1.20.1 (Forge 47.3.0 or higher)
Java 17 or higher
An LLM API key (OpenAI, Anthropic, MiniMax, DeepSeek — any OpenAI-compatible endpoint works)
Install Minecraft Forge 1.20.1 (47.3.0+)
Download hermesgod-1.0.0.jar and place it in your mods folder
Launch the game once — a config file will be created at config/hermesgod.json
Edit the config:
{
"apiEndpoint": "https://api.minimaxi.com/v1/chat/completions",
"apiKey": "your-api-key-here",
"modelName": "MiniMax-M2.7-highspeed",
"currentMode": "eco",
"chatEnabled": true
}Restart the game or server
Type /hermesgod help to see all commands
Works in single-player and multiplayer. For dedicated servers, install on the server side.
You must configure these three settings in order before the AI can respond:
| Command | Description | Example |
|---|---|---|
/hermesgod url <endpoint> | Set the API endpoint URL for your LLM provider | /hermesgod url https://api.openai.com/v1/chat/completions |
/hermesgod api <key> | Set your API key for authentication | /hermesgod api sk-xxxxxxxxxxxxxxxx |
/hermesgod model <name> | Specify which language model to use | /hermesgod model gpt-4o-mini |
| Command | Description |
|---|---|
/hermesgod help | Display all available commands |
/hermesgod mode <eco/prank/disabled> | Switch mode: eco Guardian / prank Trickster / disabled Off |
/hermesgod chat <message> | Talk to HermesGod via server console |
/hermesgod status | View current mode, API connection status, and last action |
/hermesgod interval <seconds> | Set fixed prank interval in seconds (0 to disable auto-pranks) |
/hermesgod interval <min> <max> | Set random prank interval range (in seconds) |
/hermesgod maxtokens <count> | Set max tokens per API call |
/hermesgod reset | Clear the AI's conversation memory and reset context |
Enter these three commands in order to activate the AI:
/hermesgod url https://api.openai.com/v1/chat/completions /hermesgod api sk-your-key-here /hermesgod model gpt-4o-mini
Once configured, use /hermesgod mode eco to enable Guardian mode. You can now talk to the AI in chat.
config/hermesgod.json settings:
| Setting | Default | Description |
|---|---|---|
apiEndpoint | MiniMax API | OpenAI-compatible endpoint URL |
apiKey | "" | Your API key |
modelName | MiniMax-M2.7-highspeed | Model to use |
currentMode | disabled | Default mode on startup |
chatEnabled | true | Enable chat interaction |
prankIntervalMin | 30 | Min seconds between pranks (0 = off) |
prankIntervalMax | 30 | Max seconds between pranks |
maxTokens | 8000 | Max tokens per LLM call |
temperature | 0.8 | AI creativity level |
HermesGod makes API calls to your configured LLM provider. Each call consumes tokens.
Eco mode: Triggered only by player chat messages.
Prank mode: Triggered at your configured interval (default: every 30s).
We recommend lightweight models like gpt-4o-mini or MiniMax-M2.7-highspeed to keep costs low.
Author: lql114514
Built with: Minecraft Forge 1.20.1, Gson, LLM API
License: MIT
Q: Is this a hack or cheat?
A: No. HermesGod is a server-side mod. It runs entirely on the server and gives no player unfair advantages.
Q: Can I use it on a multiplayer server?
A: Absolutely. Install on the server side. Only one API key is needed.
Q: Why does the AI sometimes not respond?
A: Check your API key, endpoint, and internet connection. All errors are logged in logs/latest.log.
Q: Can I customize the pranks?
A: Yes. The prank system is prompt-driven. Edit HermesBrain.java to customize the AI's behavior prompts.
HermesGod is watching.
HermesGod is waiting.
HermesGod is thinking — what happens next?
undefined
