Open Design MCP Server

Open Design MCP Server

Enterprise-grade MCP bridge exposing Open Design's REST API as 15 MCP tools, enabling AI agents to create, iterate, and manage design projects via the Model Context Protocol.

Category
访问服务器

README

<p align="center"> <img src="docs/screenshots/02-dashboard.png" alt="Open Design MCP Admin Dashboard" width="720"> </p>

<h1 align="center">Open Design MCP Server</h1>

<p align="center"> <strong>Enterprise-grade MCP bridge for <a href="https://opendesign.dev">Open Design</a> — manage AI design agents via the Model Context Protocol.</strong> </p>

<p align="center"> <img src="https://img.shields.io/badge/python-3.12+-blue?logo=python&logoColor=white" alt="Python 3.12+"> <img src="https://img.shields.io/badge/FastAPI-0.115+-009688?logo=fastapi&logoColor=white" alt="FastAPI"> <img src="https://img.shields.io/badge/React-19-61DAFB?logo=react&logoColor=black" alt="React 19"> <img src="https://img.shields.io/badge/MCP-Protocol-purple" alt="MCP Protocol"> <img src="https://img.shields.io/badge/Tailwind_CSS-4.0-38B2AC?logo=tailwindcss&logoColor=white" alt="Tailwind CSS 4"> <img src="https://img.shields.io/badge/License-MIT-green" alt="License MIT"> </p>

<p align="center"> <a href="README_zh-TW.md">繁體中文文件</a> </p>


Overview

Open Design MCP Server exposes the Open Design daemon's REST API as 15 MCP tools over the StreamableHttp transport, enabling any MCP-compatible AI client (Claude App, Claude Code, n8n, etc.) to create, iterate, and manage AI-generated design projects.

The package includes a full-featured Admin GUI (React 19 SPA) for monitoring, tool management, token rotation, and live log viewing — all secured with JWT authentication, CORS allowlists, and security headers.

Why This Package?

Challenge Solution
Open Design daemon only has a REST API — no MCP support FastMCP server translates 15 REST endpoints into MCP tools
MCP stdio transport isn't network-accessible Supergateway bridges stdio → StreamableHttp over HTTP
No admin interface for managing MCP tools and tokens React 19 SPA with Dashboard, Tool Manager, Token Manager
Security for exposed MCP endpoints JWT auth, CORS allowlist, SSRF protection, security headers, encrypted proxy
Long-running AI agent operations (3-5 min) SSE streaming parser with configurable timeouts (up to 300s)
Deployment complexity (3 components) Single Docker image with multi-stage build, K8s-ready

Features

Core Capabilities

  • 15 MCP Tools — System info, project CRUD, file access, AI chat (SSE streaming), content discovery
  • StreamableHttp Transport — Network-accessible MCP via Supergateway (stdio → HTTP bridge)
  • SSE Response Parser — Handles long-running AI agent operations with real-time text streaming
  • Configurable Timeouts — 60s for reads, 300s for AI chat operations

Admin GUI

  • Dashboard — Real-time health monitoring (daemon, MCP server, proxy status)
  • Tool Manager — Enable/disable individual MCP tools without restarting
  • Token Manager — Generate, rotate, and manage MCP authentication tokens
  • Log Viewer — Live-streaming server logs with severity filtering
  • Settings — Connection configuration with SSRF-protected connectivity testing

Security

  • JWT Authentication — HS256 tokens with configurable expiry (default 24h)
  • CORS Allowlist — Explicit origin allowlist (no wildcards)
  • Security Headers — HSTS, X-Frame-Options: DENY, X-Content-Type-Options, Referrer-Policy
  • SSRF Protection — Block metadata endpoints (169.254.169.254), loopback, link-local addresses
  • Request Size Limit — 1 MB body limit to prevent abuse
  • Encrypted MCP Proxy — Token-based URL path for MCP endpoint access
  • Secure Cookies — HttpOnly, Secure, SameSite=Strict

Architecture

graph TB
    subgraph External["External AI Clients"]
        Claude["Claude App / Claude Code"]
        N8N["n8n / Custom MCP Client"]
    end

    subgraph CF["Cloudflare Tunnel"]
        CFT["*.woowtech.io"]
    end

    subgraph Pod["K8s Pod: od-mcp"]
        subgraph AdminGUI["Admin GUI — FastAPI :8080"]
            Auth["JWT Auth Middleware"]
            SecHeaders["Security Headers"]
            Proxy["/private_{token}/mcp → Reverse Proxy"]
            SPA["React 19 SPA"]
            API["REST API /api/*"]
        end

        subgraph SG["Supergateway — Node.js :8000"]
            STDIO["stdio → StreamableHttp Bridge"]
        end

        subgraph MCP["MCP Server — FastMCP"]
            Tools["15 MCP Tools"]
        end
    end

    subgraph OD["Open Design Daemon"]
        ODAPI["REST API :7457"]
    end

    Claude --> CFT
    N8N --> CFT
    CFT --> Auth
    Auth --> SecHeaders
    SecHeaders --> Proxy
    Proxy --> STDIO
    STDIO --> Tools
    Tools --> ODAPI
    CFT --> SPA
    SPA --> API

Three-Component Architecture

Component Technology Port Role
MCP Server Python FastMCP stdio 15 MCP tools wrapping OD daemon REST API
Supergateway Node.js 8000 Bridges MCP stdio → StreamableHttp transport
Admin GUI FastAPI + React 19 8080 Web management console + MCP reverse proxy

Data Flow

AI Client → Cloudflare Tunnel → Admin GUI (:8080)
                                    ├── /private_{token}/mcp → Supergateway (:8000) → MCP Server (stdio) → OD Daemon (:7457)
                                    └── /api/* → REST API (settings, tools, tokens, logs)

MCP Tools (15)

Category A: System Tools (4)

Tool Description Parameters
health Check OD daemon health status and version
version Get detailed daemon version info
list_agents List available AI agent CLIs (Claude, OpenCode, BYOK)
list_connectors List external connectors (GitHub, etc.)

Category B: Project Tools (5)

Tool Description Parameters
list_projects List all design projects with status and metadata
get_project Get detailed info about a specific project project_id
create_project Create a new project via AI agent (long-running, up to 5 min) prompt, agent_id
delete_project Permanently delete a project project_id
list_project_files List all files in a project directory project_id

Category C: File Tools (2)

Tool Description Parameters
read_file Read raw content of a project file project_id, file_path
get_file_info Get file metadata (size, MIME type) project_id, file_path

Category D: AI Chat Tools (2)

Tool Description Parameters
send_message Send follow-up message in a project conversation (long-running) project_id, prompt, agent_id
list_runs List all active and completed AI agent runs

Category E: Content Tools (2)

Tool Description Parameters
list_plugins List available plugins and templates
list_skills List available design skills and triggers

Admin GUI

Login

Secure JWT-based authentication with admin password.

<p align="center"> <img src="docs/screenshots/01-login.png" alt="Login Page" width="720"> </p>

Dashboard

Real-time health monitoring — OD Daemon status, MCP Server status, Proxy status, daemon version, and namespace info.

<p align="center"> <img src="docs/screenshots/02-dashboard.png" alt="Dashboard" width="720"> </p>

Tool Manager

Enable or disable individual MCP tools. View tool categories, parameters, and descriptions.

<p align="center"> <img src="docs/screenshots/03-tools.png" alt="Tool Manager" width="720"> </p>

Token Manager

Generate and rotate MCP authentication tokens. View token history and connection URLs.

<p align="center"> <img src="docs/screenshots/04-tokens.png" alt="Token Manager" width="720"> </p>

Log Viewer

Live-streaming server logs with auto-scroll and severity filtering.

<p align="center"> <img src="docs/screenshots/05-logs.png" alt="Log Viewer" width="720"> </p>

Settings

Connection configuration, MCP server process management, and SSRF-protected connectivity testing.

<p align="center"> <img src="docs/screenshots/06-settings.png" alt="Settings" width="720"> </p>

Quick Start

Docker

# Build
docker build -t od-mcp-server:latest .

# Run
docker run -d \
  -p 8080:8080 \
  -p 8000:8000 \
  -e OD_API_BASE=http://your-od-daemon:7457 \
  -e JWT_SECRET=your-secret-key \
  -v od-mcp-data:/data \
  od-mcp-server:latest

Kubernetes (K3s)

# Apply manifests
kubectl apply -f k8s-manifests/

# Verify
kubectl get pods -n open-design
kubectl logs -n open-design deployment/od-mcp -c admin-gui

MCP Client Configuration

Connect any MCP client using the StreamableHttp endpoint:

{
  "mcpServers": {
    "open-design": {
      "url": "https://your-domain.com/private_{YOUR_TOKEN}/mcp",
      "transport": "streamable-http"
    }
  }
}

Tech Stack

Backend (Python)

Package Version Purpose
FastAPI 0.115+ Admin GUI REST API framework
Uvicorn 0.34+ ASGI server
httpx 0.28+ HTTP client for OD daemon communication
Pydantic 2.10+ Data validation and serialization
PyJWT 2.10+ JWT authentication
SSE-Starlette 2.2+ Server-Sent Events support
mcp[server] 1.1+ MCP protocol SDK (FastMCP)
PyYAML 6.0+ Configuration file parsing
aiofiles 24.1+ Async file I/O

Frontend (JavaScript)

Package Version Purpose
React 19.0 UI framework
React Router 7.0 Client-side routing
TanStack React Query 5.60 Server state management
Lucide React 0.460 Icon library
Tailwind CSS 4.0 Utility-first CSS framework
Vite 6.0 Build tool and dev server

Infrastructure

Component Purpose
Supergateway stdio → StreamableHttp MCP transport bridge
Docker (multi-stage) Node 22 (frontend build) → Python 3.12-slim (runtime)
K3s / Kubernetes Container orchestration
Cloudflare Tunnel Secure external access with TLS termination

Project Structure

.
├── od_mcp_server.py          # MCP server — 15 tools (FastMCP, stdio)
├── od_mcp_admin/             # Admin GUI app (FastAPI entry point)
│   └── routers/
│       └── config.py         # OD-specific connection config router
├── mcp_admin_core/           # Shared admin framework
│   ├── app.py                # FastAPI app factory + security middleware
│   ├── auth/
│   │   └── middleware.py     # JWT auth middleware + login router
│   ├── config.py             # JSON config store
│   ├── process.py            # MCP server process manager
│   ├── proxy.py              # MCP reverse proxy
│   └── routers/
│       └── settings.py       # Full config management API
├── frontend/                 # React 19 SPA (Vite + Tailwind CSS 4)
│   ├── src/
│   │   ├── pages/            # Dashboard, Tools, Tokens, Logs, Settings
│   │   └── components/       # Shared UI components
│   └── package.json
├── k8s-manifests/            # Kubernetes deployment manifests
├── Dockerfile                # Multi-stage build (Node 22 + Python 3.12)
├── entrypoint.sh             # Container startup script
└── pyproject.toml            # Python project configuration

Security

  • Authentication: JWT (HS256) with configurable expiry, secure cookie storage
  • Transport: All traffic encrypted via Cloudflare Tunnel (TLS 1.3)
  • CORS: Explicit origin allowlist — no wildcard *
  • Headers: HSTS, X-Frame-Options: DENY, X-Content-Type-Options: nosniff, Referrer-Policy
  • SSRF Protection: Blocks requests to cloud metadata endpoints, loopback, and link-local addresses
  • Request Limits: 1 MB body size limit
  • Token Security: MCP tokens masked in API responses, rotation with history tracking
  • API Surface: Swagger/OpenAPI docs disabled in production

License

MIT License. See LICENSE for details.


<p align="center"> Built by <a href="https://github.com/WOOWTECH">WOOWTECH</a> </p>

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
Kagi MCP Server

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。

官方
精选
Python
graphlit-mcp-server

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。

官方
精选
TypeScript
Exa MCP Server

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。

官方
精选
mcp-server-qdrant

mcp-server-qdrant

这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。

官方
精选
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选