OmniGraffle BindCraft
Enables AI agents to create publication-quality scientific diagrams, flowcharts, and architecture diagrams in OmniGraffle via natural language commands.
README
omnigraffle-bindcraft
OmniGraffle automation skill + MCP server for AI coding agents. Create publication-quality scientific diagrams, flowcharts, and architecture diagrams via natural language.
<p align="center"> <img src="https://img.shields.io/badge/platform-macOS-blue" alt="macOS only"> <img src="https://img.shields.io/badge/OmniGraffle-7.x-purple" alt="OmniGraffle 7"> <img src="https://img.shields.io/badge/MCP-compatible-green" alt="MCP compatible"> <img src="https://img.shields.io/badge/icons-2658-teal" alt="2658 icons"> <img src="https://img.shields.io/badge/license-MIT-orange" alt="MIT License"> </p>
<p align="center"> <img src="examples/demo_rag_pipeline.png" alt="Demo: RAG Pipeline" width="800"> </p> <p align="center"><em>Example output — RAG-Enhanced LLM Pipeline diagram with Nature Standard palette, smart arrow routing, and semantic color categories</em></p>
Features
- 36 MCP tools for full OmniGraffle control: shapes, connections, layout, export
- Smart arrow routing — auto-selects straight / orthogonal / curved based on shape positions; hop-over arcs at crossings
- 6 journal color palettes — Nature, Science, Cell, Lancet, Minimal, Vibrant
- 6 diagram templates — experimental workflow, data pipeline, PRISMA, system architecture, ML pipeline, clinical trial
- 2658 IconPark icons across 38 categories
- Batch operations — create dozens of shapes and connections in single calls
- Hub magnets — distribute connection points to prevent arrow pile-ups
- Dual automation — Omni Automation JS + JXA/AppleScript
Requirements
- macOS (OmniGraffle is macOS-only)
- OmniGraffle 7 (Pro recommended for full scripting support)
- Node.js >= 18
- Python 3 with
cairosvg— for icon SVG-to-PNG conversion (optional, only needed for icons)
Installation
Quick install (auto-detects runtime)
git clone https://github.com/Youn-17/omnigraffle-bindcraft.git
cd omnigraffle-bindcraft
npm install
bash scripts/install.sh
The install script auto-detects Claude Code, Codex, and WorkBuddy, then:
- Builds the project
- Adds the MCP server to each runtime's config
- Copies the skill file to each runtime's skills directory
Manual install
<details> <summary><b>Claude Code</b></summary>
npm run build
Add to ~/.claude/settings.json:
{
"mcpServers": {
"omnigraffle": {
"command": "node",
"args": ["/absolute/path/to/omnigraffle-bindcraft/dist/index.js"]
}
}
}
Install the skill:
cp .claude/skills/sci-diagram.md ~/.claude/skills/omnigraffle-bindcraft.md
</details>
<details> <summary><b>WorkBuddy</b></summary>
npm run build
Add to ~/.workbuddy/.mcp.json:
{
"mcpServers": {
"omnigraffle": {
"command": "node",
"args": ["/absolute/path/to/omnigraffle-bindcraft/dist/index.js"]
}
}
}
Install the skill:
cp .claude/skills/sci-diagram.md ~/.workbuddy/skills/omnigraffle-bindcraft.md
</details>
<details> <summary><b>Codex</b></summary>
npm run build
Add to ~/.codex/config.json:
{
"mcpServers": {
"omnigraffle": {
"command": "node",
"args": ["/absolute/path/to/omnigraffle-bindcraft/dist/index.js"]
}
}
}
</details>
<details> <summary><b>Other MCP-compatible agents</b></summary>
Any agent supporting the Model Context Protocol can use this server. Add the MCP entry to your agent's config — the format is the same across runtimes.
</details>
Set up IconPark icons (optional)
git clone https://github.com/bytedance/IconPark.git resources/iconpark
pip install cairosvg
Quick start
Once installed, just describe what you want in natural language:
Create a RAG pipeline diagram in OmniGraffle with 10 components:
User Query, Query Analyzer, Intent Router, Knowledge Retrieval,
LLM Reasoning Core, Tool Executor, Evidence Checker,
Response Generator, Safety Filter, Final Response.
Use Nature palette. Export to Desktop.
See examples/demo_rag_pipeline.md for a full walkthrough.
Tools (36)
Document & Canvas (8)
| Tool | Description |
|---|---|
get_document_info |
Get document name, path, canvases |
create_document |
Create new document (optionally from template) |
save_document |
Save current document |
list_templates |
List all available OmniGraffle templates |
create_canvas |
Create a new canvas |
list_canvases |
List all canvases |
switch_canvas |
Switch to a canvas by index |
set_canvas_size |
Set canvas dimensions |
Shapes (6)
| Tool | Description |
|---|---|
create_shape |
Create a shape with type, position, size, text, and full styling |
create_shapes_batch |
Create multiple shapes in one call |
update_graphic |
Update any property of an existing graphic |
delete_graphics |
Delete graphics by ID |
get_graphics |
List all graphics with properties |
clear_canvas |
Remove all graphics from canvas |
Connections (4)
| Tool | Description |
|---|---|
connect_shapes |
Connect two shapes with styled, auto-routed line |
connect_shapes_batch |
Batch connect with smart routing (line_type: "auto") |
create_line |
Create a standalone line between points |
set_magnets |
Set connection points on a shape for clean arrow distribution |
Layout & Organization (6)
| Tool | Description |
|---|---|
auto_layout |
Apply automatic layout |
align_graphics |
Align graphics (left/center/right/top/bottom) |
create_layer |
Create a new layer |
list_layers |
List all layers |
group_graphics |
Group graphics together |
ungroup |
Ungroup a group |
Icons (5)
| Tool | Description |
|---|---|
search_icons |
Search 2658 IconPark icons by keyword |
list_icon_categories |
List all 38 icon categories |
suggest_icons |
Get icon suggestions for a scientific context |
insert_icon |
Insert a colored icon onto the canvas |
create_node_with_icon |
Create a labeled shape with an icon inside |
Palettes & Templates (4)
| Tool | Description |
|---|---|
list_palettes |
List all 6 scientific color palettes |
get_palette |
Get a palette's full color values |
list_diagram_templates |
List all 6 diagram templates |
get_diagram_template |
Get a template's nodes, icons, and palette |
Export & Scripting (4)
| Tool | Description |
|---|---|
export_canvas |
Export to PNG, PDF, or SVG |
run_omni_js |
Execute Omni Automation JavaScript |
run_jxa |
Execute JXA code |
run_applescript |
Execute AppleScript |
Smart arrow routing
The connect_shapes_batch tool supports line_type: "auto" which automatically picks the best routing:
| Shape relationship | Auto-selected type | Why |
|---|---|---|
| Same column (Δx < 60px) | straight |
Clean vertical drop |
| Same row (Δy < 30px) | straight |
Direct horizontal |
| Cross-column & cross-row | orthogonal |
Right-angle turns |
| Target above source | curved |
Avoids overlap with forward flow |
All connections automatically get HopType.Round for clean crossing arcs.
Color palettes
| Palette | Best for | Style |
|---|---|---|
nature |
General science papers | Clean, high-contrast, colorblind-safe |
science |
Bold research graphics | Vivid but balanced |
cell |
Biology, life sciences | Elegant muted tones |
lancet |
Clinical, medical | Professional ocean blue + orange |
minimal |
Grayscale print | Maximum print compatibility |
vibrant |
Posters, presentations | High-saturation, eye-catching |
Architecture
AI Agent <── MCP Protocol ──> MCP Server (Node.js)
│
┌─────────────┼─────────────┐
│ │ │
Omni Automation JXA/AS IconPark
(evaluateJS) (osascript) (cairosvg)
│ │ │
└──────┬──────┘ SVG → PNG
│
OmniGraffle 7
Development
npm run dev # Watch mode — recompiles on changes
npm run build # One-time build
npm start # Run the server directly
License
MIT
Credits
- IconPark by ByteDance — 2658 high-quality SVG icons
- OmniGraffle by The Omni Group
- Model Context Protocol by Anthropic
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。