Motion Lab v2.0
Built for Humans & Coding Agents

Use Motion Lab as a Library,
Agent Skill, or MCP Server.

Give Claude Code, Cursor, Codex, or Antigravity direct access to all 12 motion studies, 8 interactive UI components, and 6 calibrated easing curves—via zero-install llms.txt, structured registry.json, an installable SKILL.md, or a local zero-dependency stdio MCP server (Python 3 or Node.js).

20
Live Blocks
5
MCP Tools
0
External Deps

1. Claude Code

Skill + MCP

Clone the repo once, register the zero-dependency Python 3 (or Node.js) MCP server, and symlink the Agent Skill so Claude Code applies Motion Lab tokens automatically.

Clone & Add MCP (Python 3 — zero npm/pip install)
git clone https://github.com/hmkkaii/motion-lab.git ~/motion-lab

# Add zero-dependency Python 3 stdio MCP server
claude mcp add motion-lab -- python3 ~/motion-lab/mcp/motion-lab-mcp.py

# Or use Node.js (>= 18) instead:
# claude mcp add motion-lab -- node ~/motion-lab/mcp/motion-lab-mcp.mjs

# Optional: install Agent Skill globally
mkdir -p ~/.claude/skills/motion-lab
cp ~/motion-lab/skills/motion-lab/SKILL.md ~/.claude/skills/motion-lab/SKILL.md

2. Cursor

.cursor/mcp.json

Drop this snippet into .cursor/mcp.json (project) or ~/.cursor/mcp.json (global) to call all 5 Motion Lab tools right from Cursor Agent mode.

MCP Configuration (.cursor/mcp.json)
{
  "mcpServers": {
    "motion-lab": {
      "command": "python3",
      "args": ["/ABSOLUTE/PATH/TO/motion-lab/mcp/motion-lab-mcp.py"]
    }
  }
}

3. Codex CLI

config.toml + Skill

Register Motion Lab in ~/.codex/config.toml and install the skill under ~/.codex/skills/motion-lab/SKILL.md.

~/.codex/config.toml
[mcp_servers.motion-lab]
command = "python3"
args = ["/ABSOLUTE/PATH/TO/motion-lab/mcp/motion-lab-mcp.py"]

# Install Skill for Codex:
# mkdir -p ~/.codex/skills/motion-lab
# cp ~/motion-lab/skills/motion-lab/SKILL.md ~/.codex/skills/motion-lab/SKILL.md

4. Zero-Install Web Prompt

llms.txt · registry.json

No local install needed: paste this prompt into Claude, ChatGPT, Cursor, or Antigravity so the agent fetches exact live CSS/JS code directly from GitHub Pages.

Ready-to-paste Agent Prompt
Read https://hmkkaii.github.io/motion-lab/llms.txt and fetch https://hmkkaii.github.io/motion-lab/registry.json.

Implement the "origin-menu" popover motion and the "pulse-climate" interactive dial in my project, preserving Motion Lab's exact cubic-bezier curves, asymmetric open/close durations, tabular-nums, and prefers-reduced-motion guards.
JSON-RPC 2.0 Stdio

Available MCP Tools & Resources

Both mcp/motion-lab-mcp.py and mcp/motion-lab-mcp.mjs read live from registry.json (with automatic GitHub Pages fallback).

list_motions(category?)

Returns all 12 motion studies with specs, default easing curves, triggers, and Workbench URLs. Filter by Surfaces, Navigation, Feedback, or Micro-Motion.

get_motion_spec(id)

Returns complete CSS, Vanilla JS controller, tunable CSS variables, design rationale, and reduced-motion guard for any study slug.

list_components(category?)

Returns all 8 interactive UI components with triggers, motion tokens, and live preview URLs across 6 categories.

get_component_code(id)

Returns shared widget shell CSS, component CSS, and complete self-contained HTML/JS implementation for any of the 8 widgets.

get_easing_tokens()

Returns all 6 cubic-bezier curves, 7 semantic duration tokens, ready-to-paste :root CSS block, and 5 motion principles.

Resources (3 URIs)

motion-lab://registry, motion-lab://tokens, and motion-lab://llms for instant resource mounting in MCP clients.