DeepSlate MCP
An MCP server that renders Minecraft structure files (NBT / Sponge schematic / Litematic) into PNG preview images using vanilla Minecraft resources.
README
DeepSlate MCP
An MCP (Model Context Protocol) server that renders Minecraft structure files (NBT / Sponge schematic / Litematic) into PNG preview images using vanilla Minecraft resources.
deepslate-mcp wraps the misode/deepslate
rendering engine (MIT-licensed) inside a small MCP server, so any MCP-aware
agent — including Hermes Agent —
can call a tool and get a PNG back.
Tools
| Tool | Description |
|---|---|
render_structure |
Load a .nbt/.schem/.litematic file from disk and render to PNG |
render_blocks |
Build a structure from a programmatic block list and render to PNG |
inspect_structure |
Read structure metadata (size, palette, block count) without rendering |
render_item |
Render a single Minecraft item/block as an icon |
Each tool supports configurable camera angles (isometric, top, front, side, and custom rotation), background colour, output dimensions, and zoom level.
At a glance
| Language | TypeScript / Node.js ≥ 20 |
| License | MIT (inherits upstream deepslate) |
| Upstream | misode/deepslate v0.26.0 |
| Protocol | MCP 2025-06-18 spec (stdio transport) |
| Renderer | headless WebGL via gl |
| Resources | Vanilla Minecraft assets from Mojang's client.jar |
| Tests | Vitest — 65 unit + integration tests across 9 modules |
Example call (from an MCP client)
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "render_structure",
"arguments": {
"nbt_path": "examples/fixtures/tiny_house.nbt",
"angle": "isometric",
"width": 1024,
"height": 768
}
}
}
Response includes a text summary and an inline base64-encoded PNG image.
Quick start
# Dependencies
npm install
# Build
npm run build
# Run (stdio MCP server)
node bin/deepslate-mcp.js
# Development (hot-reload via tsx)
npm run dev
# Smoke test (starts the server and exercises all tools)
npm run smoke
# Tests
npm test
# Lint (TypeScript strict check)
npm run lint
System dependencies
On Linux, headless-gl requires Mesa libraries:
apt install libgl1-mesa-dev
The server expects an X display or virtual framebuffer (Xvfb :99 -screen 0 1024x768x24).
Architecture
See ARCHITECTURE.md for the full technical design.
M0 — Foundation (scaffold, build, basic MCP server) M1 — Render core (headless WebGL, camera presets, PNG encoding) M2 — Resource pipeline (Mojang client.jar extraction, blockstate/model/texture parsing) M3 — Multi-format loaders (.nbt, .schem, .litematic) M4 — MCP tools (4 tools wrapping the pipeline) M5 — Hermes integration (MCP config, end-to-end tool calls from Hermes Agent) M6 — Hardening (error UX, cache invalidation, progress reporting, unit tests, smoke test, lint, packaging, README polish)
All milestones complete — ready for v0.1.0 publication.
🏗️ Building Design Toolkit
This repository also includes a Minecraft building design toolkit as a
Hermes Agent skill —
minecraft-building-design.
It provides:
- Format specifications for all three major Minecraft structure formats: Vanilla NBT, Create Mod schematic, and Litematica
- Python generators (zero external dependencies) to produce
.nbt,.litematicfiles from code - Design helpers (
fill_cuboid,hollow_cuboid, mirror, rotate, etc.) - Example — a modern villa generated in all three formats
To install:
# Via install script
bash skill/install.sh
# Or manually copy
cp -r skill/minecraft-building-design ~/.hermes/skills/creative/
Then in Hermes:
skill_view(name='minecraft-building-design')
The skill complements deepslate-mcp perfectly:
- deepslate-mcp renders existing
.nbt/.schem/.litematicfiles to PNG - minecraft-building-design builds those files from scratch
See skill/minecraft-building-design/SKILL.md
for full documentation.
Resource caching
Vanilla resources (~3,500 blockstates, ~10,000 models, ~1,500 textures) are extracted from Mojang's client.jar and cached to disk on first build. Subsequent loads take ~0.2s instead of ~7s. Purge the cache with:
rm -rf ~/.cache/deepslate-mcp
Environment variables
| Variable | Purpose |
|---|---|
DEEPSLATE_JAR_PATH |
Point to a pre-downloaded client.jar (skips network) |
DEEPSLATE_CACHE_DIR |
Override the cache directory (default: ~/.cache/deepslate-mcp) |
DISPLAY |
X display for headless-gl (e.g. :99) |
Credits
This project is built on top of
misode/deepslate by Misode,
licensed under MIT. See ATTRIBUTION.md for full
upstream credits and license obligations.
Minecraft content and assets are © Mojang AB; this project never distributes Mojang-owned assets — all resources are fetched at runtime from official sources.
License
MIT — see LICENSE.
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.