AI Charts
An MCP server that enables AI assistants to programmatically create and manage flowcharts, ERDs, and swimlane diagrams. It provides tools for manipulating diagram structures, performing auto-layouts, and exporting to Mermaid or Markdown formats.
README
<p align="center"> <img src="ai-charts-header.png" alt="AI Charts" width="100%" /> </p>
AI Charts
AI-powered flowchart, ERD, and swimlane diagram builder.
AI Charts combines a built-in AI assistant for quick chart creation with an MCP server that lets external AI tools like Claude Desktop and Cursor create and manage charts programmatically. It works with any OpenAI-compatible LLM provider — Ollama, OpenAI, or your own endpoint. No vendor lock-in.
Features
- Three chart types — flowcharts, entity-relationship diagrams, and swimlane diagrams
- Built-in AI assistant — setup wizard for new projects and a chart builder mode for iterating on diagrams through conversation
- MCP server — 18+ tools for external AI integration; connect Claude Desktop, Cursor, or any MCP client to create and manage charts programmatically
- Any LLM provider — works with Ollama, OpenAI, or any OpenAI-compatible API
- Import/export — portable ZIP project archives, Mermaid syntax, Markdown summaries, and PDF export
- Auto-layout & validation — ELK-based auto-layout, cycle detection, reachability analysis, and structural validation
- Revision history — full undo/redo with event tracking
- Dark-themed UI — built with React Flow for interactive diagram editing
Quick Start
Prerequisites: Bun and an LLM provider (e.g. Ollama)
git clone https://github.com/tjameswilliams/ai-charts.git
cd ai-charts
bun install
bun run db:push
bun run dev
The app will be available at http://localhost:5174.
LLM Configuration
Configure your LLM provider through the Settings UI in the app. The default configuration points to a local Ollama instance:
| Setting | Default |
|---|---|
| API Base URL | http://localhost:11434/v1 |
| API Key | ollama |
| Model | llama3.2 |
To use OpenAI, set the base URL to https://api.openai.com/v1, add your API key, and choose a model like gpt-4o. Any OpenAI-compatible endpoint works the same way.
MCP Server
AI Charts exposes an MCP server so external AI tools can create and manage charts programmatically.
Run standalone:
bun run mcp
Connect from Claude Desktop — add to your claude_desktop_config.json:
{
"mcpServers": {
"ai-charts": {
"command": "bun",
"args": ["run", "mcp"],
"cwd": "/path/to/ai-charts"
}
}
}
Available tools include: list_projects, create_project, list_charts, create_chart, build_chart, add_node, add_edge, delete_node, delete_edge, resize_nodes, get_nodes, get_chart_status, validate_chart, export_mermaid, export_markdown, and more.
Project Structure
client/ React + Vite frontend
src/
components/ UI components
store/ Zustand state management
api/ API client
server/ Bun + Hono backend
routes/ API endpoints
db/ Drizzle ORM schema & migrations
lib/
llm.ts LLM integration & tool definitions
mcp/ MCP server & client manager
export/ Export utilities (Mermaid, Markdown, PDF)
validation.ts Chart validation
Tech Stack
- Runtime: Bun
- Backend: Hono, SQLite via Drizzle ORM
- Frontend: React 19, React Flow, Zustand, Tailwind CSS 4
- AI: OpenAI-compatible chat completions, Model Context Protocol
- Layout: ELK via elkjs
License
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。