blockbench-mcp-modeling

Blockbench MCP Modeling Operator Guide

This skill makes you good at driving Blockbench through the BlockbenchMCP bridge (mcp__blockbench__* tools). It encodes a modeling doctrine and a set of scripts verified live against Blockbench 5.1.4 + GeckoLib.

The failure this skill exists to prevent: you cannot hold a 3D coordinate system in your head, so when asked for a hooded warrior you emit one box for the torso, one for the cloak and one for the head — 15 monoliths, no layers, no silhouette. That is a tooling problem, and the tooling now exists. Your job is the SHAPE and the LAYERS; the generators do the arithmetic.

Connection check (do this first)

The tools talk to a local server inside Blockbench. If a tool returns Cannot reach Blockbench on http://127.0.0.1:8787:

• The desktop Blockbench app must be open and the MCP server started: in Blockbench, Tools ▸ Start MCP Server, or the MCP Copilot side panel's Start button. The server stops when Blockbench is closed.
• Then call get_status to confirm. Before building, read get_guide {topic:"detailing"}.
• The MCP Copilot panel is where the user drops a reference image, watches your activity log, and sees the live match score. If they mention a reference, call get_reference.

1. Cube budget — the non-negotiable floor

What you are building	Cubes
Simple prop / small item (potion, key, dagger, crate)	30–60
Standard mob / NPC / creature	100–180
Hero model / boss / detailed character	180–300+

A humanoid built from fewer than 70 cubes is a draft, not a model. It does not go to texturing, it does not go to review, it gets more geometry. audit_complexity is the gate and returns too_primitive / acceptable / high_detail — treat too_primitive as an instruction, not an opinion.

Cube count is necessary, not sufficient: 200 cubes in one flat plane is still flat. The budget buys you the four layers below.

2. The 4-layer doctrine — build outward