Minecraft MCP Server Pro
Enables AI agents to control Minecraft via MCP with configurable versions, creative building commands, survival helpers, and an autonomous agent loop.
README
Minecraft MCP Server Pro
Fork of yuniko-software/minecraft-mcp-server upgraded into a layered platform for AI control:
- Configurable Minecraft version (default
1.20.1, not hard-pinned) - Offline / LAN login
- Command channel (
/fill,/setblock, batch files) for fast creative builds - Survival helpers (status / collect / eat; craft tools from upstream)
- Autonomous agent loop (
agent-start) — plan → skills → observe
Design: ../docs/superpowers/specs/2026-07-24-minecraft-mcp-pro-design.md
Prerequisites
- Node.js >= 20.10.0
- Minecraft Java LAN world (offline OK)
- Cheats enabled for
/filltools - MCP client (ZCode, Claude Desktop, …)
Install & run
cd minecraft-mcp-server
npm install
npm run build
npm start -- --username TraeBot --version 1.20.1 --host localhost --port 25565
Dev (tsx):
npm run dev -- --username TraeBot --version 1.20.1
ZCode MCP config example
{
"minecraft": {
"type": "stdio",
"command": "node",
"args": [
"./dist/main.js",
"--host", "localhost",
"--port", "25565",
"--username", "TraeBot",
"--version", "1.20.1"
],
"timeoutMs": 999999
}
}
Optional: point --batch-root at your workspace so run-command-batch can load minecraft-city-build/command_batches/*.txt:
--batch-root "."
CLI flags
| Flag | Default | Meaning |
|---|---|---|
--host |
localhost | LAN host |
--port |
25565 | LAN port |
--username |
LLMBot | Offline name |
--version |
1.20.1 | Mineflayer protocol version (change freely) |
--auth |
offline | offline | microsoft (v1: use offline) |
--capabilities |
all | baseline,command,build,survival,agent |
--batch-root |
parent of cwd | Sandbox for batch file paths |
--action-delay-ms |
50 | Delay between commands |
--no-agent |
— | Disable agent tools |
--max-agent-steps |
50 | Agent loop limit |
New / enhanced tools
Command
run-chat-command— whitelisted chat commandsfill-region— auto-split ≤32768setblock-atclone-region
Build
build-box— solid / hollow / outlineclear-regionrun-command-batch— array or file under batch-rootbuild-status
Survival / storage / social
get-player-status/collect-blocks/eat-food/ensure-itemsleep-nearby— 附近床睡觉过夜chest-deposit/chest-withdraw/chest-deposit-allfind-player/go-to-player/follow-player/fight-nearby/world-time-statusqueue-reset— 卡死队列强制恢复- (+ upstream
craft-item/list-recipes/ …)
Blueprint
build-blueprint— JSON 蓝图(blocks[]或palette+layers),见examples/blueprints/mini-hut.json
Agent (autonomous, near-human) — package 3.11.0
agent-start— NL goal,如附近搭石屋/建高塔/盖别墅/修城堡/建灯塔/跟着我/护卫我/ …agent-companion— 持续队友模式(靠近→注视→跟随→拾取→探索 循环,直到agent-cancel)agent-status/agent-memory/agent-wait/agent-cancel/agent-chat-control- 游戏内聊天也可驱动:搭石屋/建塔/别墅/城堡/灯塔/修路/平整/巡逻/护卫 等
Powerful builds (3.11):
- MCP:
build-structure+list-structures— templates: tower, villa, castle, lighthouse, pyramid, arena, fountain, road, stairs, platform, wall, hut - Agent skill:
build-structure(same templates, near player) - Low-level still:
fill-region/build-box/build-blueprint/run-command-batch
Agent skills (40+): … + build-structure, pick-up, flatten, patrol, place-bed, pillar-up, clear-lava, mark/go-waypoint
Pathing / creative shortcuts: fly/tp salvage; gather/craft/ensure/minekit creative /give
MCP tools: 54 (incl. build-structure, list-structures, agent-memory)
Reload ZCode MCP after rebuild →
pro=3.11.0-dist. Only one TraeBot.
Memory: batch-root/.minecraft-mcp-memory/<bot>.json 记住家坐标等(跨重启)
Reload ZCode MCP after rebuild so the live bot process loads new
dist/(agent plans otherwise stay on old skills). Only one TraeBot process — do not also runnode dist/main.jsin a terminal.
Offline smoke (before opening the game)
npm run smoke
# or with LAN already open:
npm run smoke -- --host localhost --port 25565 --version 1.20.1
Blueprint example
# after MCP connected + cheats on:
# build-blueprint filePath=examples/blueprints/mini-hut.json
# (batch-root should include this repo, e.g. the repo path or workspace)
Example prompts
Fast creative build
用 fill-region 在我附近清空一块 20x20 地面,再用 build-box 搭一个 10x10x6 白混凝土空心房。
Batch from city builder
run-command-batch filePath=
minecraft-city-build/command_batches/xxx.txt(需设置 batch-root)
Autonomous loop
agent-start goal=
在附近搭石屋,先保证自己不饿
Version policy
Version is configurable. Default validated target: 1.20.1.
If join fails, try matching your world version: --version 1.20.4 etc.
Safety
Default command blacklist: op, stop, ban, …
Default reject: kill @a / clear @a.
Batch files cannot escape --batch-root.
License
Apache-2.0 (upstream). Pro changes in this fork under the same license terms.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.