Skip to main content

VRChat Recipes

This guide introduces "recipes" that bundle commonly used VRChat production workflows. A recipe combines multiple tool calls into a single procedure that can be executed with a single natural language command.

What Are Recipes​

Recipes define a series of operations in a YAML file for reuse. For example, "setting up a VRM as a VRChat avatar" normally requires 5-8 tool calls, but with a recipe, it completes in a single command.

Benefits of recipes:

  • Time savings: Execute multiple steps at once
  • Standardized procedures: Achieve consistent setup quality every time
  • Easy sharing: Share YAML files across a team
  • Customizable: Adapt to different configurations by changing variables

Available VRChat Recipes​

avatar-quick-setup (Quick Avatar Setup)​

A recipe that sets up a VRChat avatar from a VRM file in one go.

Steps performed:

  1. Import VRM file into Unity
  2. setup_vrc_avatar to configure VRChat avatar settings
  3. set_viewpoint to auto-configure the viewpoint
  4. setup_hair_physbone to set up hair PhysBone
  5. create_expression_preset to create 8 basic expressions
  6. add_gesture_expression for gesture mapping
  7. check_avatar_rank to check performance rank
  8. validate_for_upload for pre-upload validation

How to run:

You:

Run the avatar quick setup recipe. The VRM file is "my_avatar.vrm".

AI action: The 8 steps above run in sequence, with results reported together.

AI response example:

Avatar quick setup complete:

1. VRM import: OK (my_avatar.vrm -> Assets/Avatars/my_avatar/)
2. Avatar settings: OK (VRC_AvatarDescriptor configured)
3. Viewpoint: OK (auto-detected and placed)
4. PhysBone: OK (24 hair bones configured)
5. Expressions: OK (8 expression presets created)
6. Gestures: OK (7 gestures mapped to expressions)
7. Rank: Medium (4 materials, 138 bones)
8. Validation: OK (ready to upload)

Time elapsed: approximately 45 seconds

world-basic-setup (Basic World Setup)​

A recipe that sets up the basic structure of a VRChat world in one step.

Steps performed:

  1. setup_vrc_world to configure the world descriptor
  2. Create a floor object with colliders
  3. add_spawn_point to place the spawn point
  4. create_light to add a directional light
  5. Add a point light
  6. add_vrc_mirror to place a mirror
  7. validate_for_upload for validation

How to run:

You:

Run the basic world setup recipe.

AI action: The basic world structure is automatically set up.

AI response example:

Basic world setup complete:

1. World descriptor: OK (VRC_SceneDescriptor configured)
2. Floor: OK (50m x 50m with colliders)
3. Spawn point: OK (placed near origin)
4. Directional light: OK (angle 50 degrees, white light)
5. Point light: OK (placed above spawn point)
6. Mirror: OK (2m x 2.5m)
7. Validation: OK (ready to upload)

Add objects and decorations from this base to build out your world.

outfit-change (Outfit Switching)​

A recipe that automates Modular Avatar outfit switching setup.

Steps performed:

  1. Place outfit Prefab as a child of the avatar
  2. auto_setup_outfit for automatic setup
  3. add_ma_toggle to add a toggle to the Expression Menu
  4. add_ma_blendshape_sync to sync body shape BlendShapes
  5. validate_for_upload for validation

How to run:

You:

Run the outfit switching recipe. The outfit is "SummerDress" and the toggle name is "Summer Dress".

AI action: Outfit setup and menu registration run automatically.

AI response example:

Outfit switching setup complete:

1. Placement: OK (SummerDress placed as child of avatar)
2. Auto setup: OK (MA Merge Armature, body mesh hiding configured)
3. Toggle: OK ("Summer Dress" added to Expression Menu)
4. BlendShape sync: OK (3 body shape keys synced)
5. Validation: OK

Verify switching from the Expression Menu.

avatar-optimize (Avatar Optimization)​

A recipe that improves avatar performance rank.

Steps performed:

  1. check_avatar_rank to check current rank
  2. suggest_optimizations to get optimization suggestions
  3. auto_optimize_avatar to run automatic optimization
  4. check_avatar_rank to check post-optimization rank
  5. compare_before_after to display before-and-after comparison

How to run:

You:

Run the avatar optimization recipe. Target Good rank.

AI action: From checking the current rank through optimization to result comparison, everything runs automatically.

AI response example:

Avatar optimization complete:

1. Pre-optimization rank: Poor
- Material count: 6 (Poor)
- Mesh count: 5 (Medium)
- Bone count: 142 (Medium)

2. Optimization suggestions: material merging, mesh merging, bone reduction

3. Auto optimization executed:
- Texture atlas created -> Materials 6 -> 1
- Mesh merging -> 5 -> 1
- Unused bones removed -> 142 -> 128

4. Post-optimization rank: Good

5. Comparison:
Materials: 6 -> 1 (Poor -> Excellent)
Meshes: 5 -> 1 (Medium -> Excellent)
Bones: 142 -> 128 (Medium -> Good)
Overall: Poor -> Good

expression-setup (Expression Setup)​

A recipe that automates FaceEmo expression setup.

Steps performed:

  1. auto_detect_blendshapes to detect available BlendShapes
  2. create_expression_preset to create 8 basic expressions
  3. add_gesture_expression for gesture mapping
  4. Add an expression submenu to the Expression Menu

How to run:

You:

Run the expression setup recipe.

AI action: From BlendShape detection through expression creation to menu configuration, everything runs automatically.

AI response example:

Expression setup complete:

1. BlendShape detection: 42 found (mouth: 19, eyes: 8, brows: 6, other: 9)
2. Expression presets: 8 expressions created
- Smile, Angry, Sad, Surprised, Shy, Wink, Jitome, Tongue
3. Gesture mapping:
- Fist -> Angry
- Victory -> Smile
- HandOpen -> Surprised
- FingerPoint -> Wink
- RockNRoll -> Tongue
- ThumbsUp -> Shy
4. Expression Menu: 8 items added to "Expressions" submenu

How to Run Recipes​

Run with Natural Language​

The simplest approach is to give natural language instructions.

"Run the avatar quick setup recipe"
"Start with the basic world setup recipe"
"Run the expression setup recipe on my_avatar"

The AI calls the run_recipe tool to execute the recipe.

Run from the Command Line​

# List available recipes
uvx openforge-mcp --list-recipes

# Run a recipe
uvx openforge-mcp --run-recipe avatar-quick-setup \
--var vrm_file="my_avatar.vrm"

# Run with variables
uvx openforge-mcp --run-recipe world-basic-setup \
--var floor_size=100 \
--var spawn_count=4

View Available Recipes​

You:

List the available VRChat recipes.

AI response example:

VRChat recipe list:
1. avatar-quick-setup - VRM -> VRChat avatar full setup
2. world-basic-setup - Basic world structure setup
3. outfit-change - Modular Avatar outfit switching
4. avatar-optimize - Performance rank optimization
5. expression-setup - FaceEmo expression setup

Creating Custom Recipes​

You can define your own workflows as recipes.

Creating a Recipe File​

Recipes are written in YAML format. Below is an example of a custom recipe that applies specific PhysBone settings to an avatar.

name: my-physbone-setup
description: Apply my preferred PhysBone settings to an avatar
version: "1.0"
author: your-name

variables:
hair_spring: 0.25
hair_stiffness: 0.3
skirt_spring: 0.4
skirt_stiffness: 0.2

steps:
- name: Set up hair PhysBone
tool: setup_hair_physbone
params:
preset: "long"
spring: "{{ hair_spring }}"
stiffness: "{{ hair_stiffness }}"

- name: Set up skirt PhysBone
tool: setup_skirt_physbone
params:
spring: "{{ skirt_spring }}"
stiffness: "{{ skirt_stiffness }}"

- name: Add left leg collider
tool: add_physbone_collider
params:
target: "LeftUpperLeg"
type: "Capsule"
radius: 0.04

- name: Add right leg collider
tool: add_physbone_collider
params:
target: "RightUpperLeg"
type: "Capsule"
radius: 0.04

- name: Take confirmation screenshot
tool: take_screenshot
params:
width: 1920
height: 1080

Recipe File Locations​

LocationPurpose
./recipes/Project-local (specific to that project)
~/.openforge/recipes/User-wide (available across all projects)

You:

Save this recipe to the project.

AI action: The YAML file is saved as ./recipes/my-physbone-setup.yaml.

Recipe Variables​

Recipe variables can be overridden at runtime.

You:

Run the my-physbone-setup recipe. Set the hair spring to 0.5.

AI action: The recipe runs with the hair_spring variable overridden to 0.5.

Conditional Steps​

You can branch processing based on the result of a previous step.

steps:
- name: Check performance rank
tool: check_avatar_rank
params: {}

- name: Run optimization if needed
tool: auto_optimize_avatar
params:
target_rank: "Good"
condition: "{{ previous.rank != 'Good' and previous.rank != 'Excellent' }}"

- name: Show results
tool: compare_before_after
params: {}
condition: "{{ previous.optimized == true }}"

Error Handling​

Specify what happens when a step fails.

steps:
- name: Detect BlendShapes
tool: auto_detect_blendshapes
params: {}
on_error: stop # Stop the recipe on error (default)

- name: Create expression presets
tool: create_expression_preset
params: {}
on_error: continue # Proceed to next step even on error

Recipe Use Cases​

Team Standardization​

Place recipes in the project's recipes/ directory so the entire team uses the same setup procedures.

my-vrchat-project/
recipes/
team-avatar-setup.yaml
team-world-setup.yaml
team-optimization.yaml
Assets/
...

Automating Repetitive Tasks​

Useful when applying the same settings to multiple avatars.

You:

Run the my-physbone-setup recipe on all avatars in the scene.

AI action: The recipe runs on all objects in the scene that have a VRC_AvatarDescriptor.

Next Steps​

Use recipes to efficiently create VRChat content.