Architecture Flow (more or less)
Website *Command Center*. Hooks up to our custom API to connect to the server and coordinate between different backend services.
AI agents socializing with eachother (and giving items)
the bots deadlocked or something....

E-Steves

STOP exploiting villagers, START using AI.

Inspiration

Minecraft automation has always come at a cost — usually in the form of endlessly exploited villagers, iron farms, and morally questionable gameplay loops. We asked ourselves: what if we didn’t have to torture villagers just to get resources?

E-Steves reimagines Minecraft labor: AI agents do the work so players can focus on creativity, exploration, and strategy.

What It Does

E-Steves allows players to deploy AI-powered bot players that autonomously gather materials, perform tasks, and follow high-level instructions on a Minecraft server — even while the player is offline.

How We Built It

E-Steves is built on top of a programmable Minecraft bot framework.

At a high level:

Node.js manages Minecraft bot instances and server communication
Each agent runs as an independent, modular process
Agents can perform actions such as movement, block interaction, and resource gathering
A web-based interface allows users to command, manage, and monitor agents remotely
The system supports multiple concurrent agents for scalable automation

Challenges We Solved

LLM endpoint integration: Designing a reliable LLM endpoint that could interpret server logs and translate high-level instructions into actionable agent behaviors was non-trivial.
Instruction transfer: Converting natural language commands into structured, executable actions for Minecraft agents required careful parsing and validation.
Server log handling: Streaming, filtering, and interpreting Minecraft server logs in real time without overwhelming the LLM or introducing latency was a major challenge.
Agent synchronization: Keeping multiple agents aligned with updated instructions.