MCPInstall

blockbench-hytale

Create Hytale models and animations through Blockbench MCP. Use for character/attachment or prop formats, UV density, stretch, quads, shading, attachment pieces, and visibility keyframes. Requires the Hytale Models plugin.

Installs
836
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/jasonjgardner/blockbench-mcp-project --skill blockbench-hytale
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads blockbench-hytale automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Blockbench Hytale

Use the shared modeling, texturing, and animation workflows where applicable.

Select the Actual Hytale Format

Call get_capabilities: include_tools=true and choose a returned Hytale format ID. Creating a Bedrock project does not activate Hytale.

FormatUnits per world blockUse
hytale_character64Characters and attachments, including tools, weapons and cosmetics
hytale_prop32World props, furniture and blocks

These numbers describe geometry/texel density, not maximum atlas resolution. Texture width and height may differ and must each be multiples of 32. Match effective UV dimensions to the actual image, and preserve the format's dimension-linked face UVs. The official Hytale art guide explains density and atlas sizing.

get_capabilities: format_id="hytale_character"
create_project: name="goblin", format="hytale_character"
hytale_get_format_info
# → formatType, blockSize, animationFPS, nodeCount, maxNodes, nodeCountValid, features

Use create_project only when a new project is intended. Inspect existing projects before replacing content.

For a new static 128×96 Hytale atlas, explicitly match UV and bitmap dimensions (both Hytale formats support per-texture UV sizing):

create_texture: name="goblin_atlas", width=128, height=96,
  uv_width=128, uv_height=96, fill_color="#808080", layer_name="base"
list_textures
# Confirm uv_size, bitmap_size and frame_size are [128,96]

The uv_width/uv_height pair is available in the current source plugin; confirm the loaded tool schema supports it. Creating this image does not pack the model's UVs; follow the texturing skill's native template workflow when a new layout is needed.

Geometry and UVs

The format supports cubes/quads, joint-based groups, stretch, optional box UV and per-texture UV dimensions. Use integer base dimensions when following the plugin's default size setting; stretch allows finer visible dimensions while retaining the UV layout. Fractional base dimensions are not categorically impossible, but make UV consistency harder. Exported models have a 255-node limit; exporter folding of a group's main cube affects the actual count. These are format behaviors documented in the Hytale format source.

add_group: name="rig_root", origin=[0, 0, 0]
add_group: name="body", parent="rig_root", origin=[0, 24, 0]
place_cube: elements=[{name: "torso_geo", from: [-8, 24, -4], to: [8, 48, 4]}], group="body"
hytale_set_cube_stretch: cube_id="torso_geo", stretch=[1.1, 1, 1]
hytale_get_cube_stretch: cube_id="torso_geo"

hytale_create_quad: name="leaf", position=[0, 16, 0],
  normal="+Y", size=[8, 8], double_sided=true

Normals may be +X, -X, +Y, -Y, +Z, or -Z. Use double-sided rendering when both sides should appear. Small stretch changes can fix intersections; the art guide recommends roughly 0.7–1.3 per axis for its style, which is a visual recommendation rather than the MCP schema's validity range. Block in colors before painted volume and fine detail, and avoid unintended noisy grain. Soft brushes are appropriate to Hytale's painted style; pixel-perfect drawing is not mandatory. See the art guide.

set_cube_uv preserves Hytale's autouv=1 and dimension-linked rectangle sizes. Move or mirror a face's UVs while retaining its absolute base geometry width/height (swapped for 90/270-degree UV rotation); changing those extents is rejected before Undo. Stretch changes the visible shape, not that base UV size. For the per-face torso_geo above, its unrotated north face is 16×24:

get_cube_uv: id="torso_geo"
set_cube_uv: id="torso_geo", faces={north: {uv: [8, 8, 24, 32], rotation: 0}}

Hytale quads cannot convert to box UV. Texture selection follows an attachment collection or the project default, even when the format reports single_texture=false; non-null per-face texture assignments are rejected. Inspect effective_texture and use the native collection/default texture workflow. See native element behavior and attachment texture resolution.

Shading

hytale_set_cube_properties sets shading_mode and double_sided; hytale_get_cube_properties reads them. Accepted shading names are standard, flat, fullbright, and reflective. The exported property is not proof of matching viewport appearance; verify materials in Hytale.

hytale_set_cube_properties: cube_id="torso_geo", shading_mode="standard"
hytale_get_cube_properties: cube_id="torso_geo"

Attachments

hytale_list_attachments lists existing attachment collections. hytale_set_attachment_piece marks an existing group; hytale_list_attachment_pieces reads those markers. A piece attaches to a matching base-model bone name, so inspect both assets and preserve the required names.

# Existing attachment group whose name matches the intended base bone
hytale_set_attachment_piece: group_name="hand_right", is_piece=true
hytale_list_attachment_pieces
hytale_list_attachments

Marking a group does not create a collection, configure the game's item, or export a separate attachment. The current tool family lists collections but does not create them. Use the installed plugin's native collection/attachment workflow when needed. The model compiler expects a live Collection for attachment-specific export; a JSON string in generic export_model.options is not that object. Verify the attachment with the base model and inspect the actual exported nodes. See Hytale plugin features and model codec source.

Animation

Hytale animation files use 60 frame units per second. MCP keyframe times are seconds; time=0.5 represents frame 30 at export. Use the animation UUID returned by create_animation and existing group names:

draw = create_animation: name="draw", animation_length=1, bones={}
hytale_create_visibility_keyframe: animation_id=draw.uuid,
  bone_name="weapon_sheathed", time=0, visible=true
hytale_create_visibility_keyframe: animation_id=draw.uuid,
  bone_name="weapon_sheathed", time=0.5, visible=false
hytale_set_animation_loop: animation_id=draw.uuid, loop_mode="hold"

The examples use tool-call pseudocode; bind draw.uuid to the returned UUID. Visibility changes need initial states as well as transition keys. For swapping two representations, key both at the same time.

Hytale uses quaternion interpolation and wraps looping clips. For full turns use intermediate orientations with less than 180 degrees between keys, and preview the direction and loop seam. Native UV-offset channels exist, but the current MCP family has no dedicated UV-offset-keyframe tool.

The upstream .blockyanim exporter writes holdLastKeyframe, and maps Catmull-Rom to smooth and other interpolation to linear. The editor's once, step or Bezier state therefore does not establish equivalent runtime playback. Verify the actual export and target behavior. These details come from the animation codec source.

Validation and Export

  1. Inspect UV/bitmap dimensions, texture assignments, stretch and geometry in relevant poses.
  2. Run hytale_validate_model, then inspect its issues and scope. The current implementation counts nodes in the installed blockymodel compiler's main-model output and checks that both texture dimensions are positive multiples of 32. It reports main_model_node_count_and_texture_dimensions; attachments and full engine compatibility need separate checks. Older bundles used heuristic node counts and incorrect density-based texture checks, so discover/reload the current build before relying on this result.
  3. Discover list_export_formats; use the registered blockymodel codec for an intended model export. The texture image is a separate deliverable.
  4. Export animations through the installed plugin's native animation workflow. Upstream provides the export_blockyanim action; do not invent a blockyanim model codec when discovery does not return one. Complete any resulting native save dialog through the authorized UI workflow.
  5. Reopen a copy or inspect the runtime asset when available. Report whether texture, attachment, shading, interpolation and loop behavior were actually checked in Hytale.

Resources include hytale://format, hytale://attachments/{id}, hytale://pieces/{id}, and hytale://cubes/{id}. Technical references checked 2026-09-13; the running Hytale plugin version remains authoritative for available behavior.

FAQ

What does the blockbench-hytale skill do?

Create Hytale models and animations through Blockbench MCP. Use for character/attachment or prop formats, UV density, stretch, quads, shading, attachment pieces, and visibility keyframes. Requires the Hytale Models plugin.

How do I install blockbench-hytale?

Run npx skills add https://github.com/jasonjgardner/blockbench-mcp-project --skill blockbench-hytale in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/blockbench-hytale/ (personal) or .claude/skills/blockbench-hytale/ (project).

Which AI agents can use blockbench-hytale?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is blockbench-hytale safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains blockbench-hytale?

It is published by jasonjgardner in jasonjgardner/blockbench-mcp-project. It has 836 recorded installs.

More from jasonjgardner/blockbench-mcp-project

View all →

blockbench-modeling

jasonjgardner/blockbench-mcp-project

Create and edit 3D models in Blockbench using MCP tools. Use when building geometry with cubes, creating meshes, placing spheres/cylinders, editing vertices, extruding faces, or organizing models with groups. Covers both cube-based Minecraft modeling and freeform mesh editing.

AI & Agents1.3K installs

blockbench-texturing

jasonjgardner/blockbench-mcp-project

Create and paint textures in Blockbench using MCP tools. Use when creating textures, painting on models, using brush tools, filling colors, drawing shapes, applying gradients, managing texture layers, or working with UV mapping. Covers pixel art texturing, procedural painting, and UV manipulation.

AI & Agents1.2K installs

blockbench-animation

jasonjgardner/blockbench-mcp-project

Create and manage animations in Blockbench using MCP tools. Use when animating 3D models, creating keyframes, managing bone rigs, editing animation curves, or working with animation timelines. Covers walk cycles, idle animations, combat animations, and complex multi-bone animations.

Frontend & Design1.1K installs

blockbench-mcp-overview

jasonjgardner/blockbench-mcp-project

Discover and use Blockbench MCP tools, resources, and prompts. Use for capability questions, new-project planning, mode navigation, export, display transforms, Bedrock face materials, or native armature workflows.

AI & Agents1.1K installs

blockbench-use

jasonjgardner/blockbench-mcp-project

Load before Blockbench MCP calls that create, modify, or export content. Discover the active format and tools, route to modeling, UV/texturing, animation, particle, PBR, or Hytale guidance, and verify the requested result.

AI & Agents905 installs

blockbench-pbr-materials

jasonjgardner/blockbench-mcp-project

Create and manage PBR materials in Blockbench using MCP tools. Use for Bedrock textureset.json import/export, normal or height maps, packed MER textures, material channel replacement, and uniform material properties.

AI & Agents861 installs

Popular ai & agents skills

View category →

google-agents-cli-publish

google/agents-cli

This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).

AI & Agents513.9K installs

google-agents-cli-eval

google/agents-cli

This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).

AI & Agents513.5K installs

skill-creator

anthropics/skills

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

OfficialAI & Agents416.4K installs

codex-pet

prime-skills/runcomfy-agent-skills

Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.

AI & Agents402.1K installs

full-output-enforcement

leonxlnx/taste-skill

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

AI & Agents401K installs

executing-plans

obra/superpowers

Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available

AI & Agents245.7K installs