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
Visit Server

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>

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured