Maps MCP App
Enables users to search for locations, businesses, and points of interest on an interactive OpenStreetMap map within ChatGPT conversations.
README
Maps MCP App
A ChatGPT App that displays an interactive map powered by OpenStreetMap data. Users can search for locations, businesses, landmarks, and points of interest — all within a ChatGPT conversation.
Architecture
ChatGPT
↓
ChatGPT App (React + MapLibre)
↓
MCP Protocol
↓
MCP Server (Hono + Cloudflare Workers)
↓
OpenStreetMap API (Nominatim + Overpass)
Tech Stack
| Layer | Technology |
|---|---|
| Frontend | React 19, TypeScript, Vite, MapLibre GL JS |
| Backend | Bun, Hono, MCP SDK, Zod |
| Data | OpenStreetMap (Nominatim geocoding, Overpass POI queries) |
| Deployment | Cloudflare Workers (server) + Cloudflare Pages (frontend) |
| Validation | Shared Zod schemas across frontend and backend |
Repository Structure
├── apps/
│ ├── mcp-server/ # MCP protocol server
│ │ └── src/
│ │ ├── routes/ # Hono HTTP routes
│ │ ├── tools/ # MCP tool handlers
│ │ ├── services/ # OSM search service
│ │ └── index.ts # Cloudflare Worker entry
│ └── chatgpt-app/ # React frontend
│ └── src/
│ ├── components/ # React components
│ ├── context/ # State management
│ ├── api/ # MCP client
│ └── main.tsx # App entry
├── packages/
│ ├── shared-types/ # TypeScript types, enums, interfaces
│ └── shared-validation/ # Zod schemas (shared)
└── docs/
├── techspecs.md # Technical specification
├── implementation.md # Implementation guide
└── DEPLOYMENT.md # Deployment instructions
Getting Started
Prerequisites
- Bun 1.3+
- Node.js 22+ (for tooling)
Install
git clone <repo-url>
cd maps-mcp-app
bun install
Build Shared Packages
bun run --filter @maps-mcp/shared-types build
bun run --filter @maps-mcp/shared-validation build
Run Locally
Terminal 1 — MCP Server:
cd apps/mcp-server
bun run dev
Server starts on http://localhost:3001
Terminal 2 — Frontend:
cd apps/chatgpt-app
bun run dev
App opens at http://localhost:5173
Run Tests
# All tests
bun test
# Or per package
cd packages/shared-validation && bun test
cd apps/mcp-server && bun test
MCP Tools
search_locations
Search for nearby points of interest.
Input:
center—{ lat: number, lng: number }radius— meters (250, 500, 1000, 2000, 5000, 10000, 25000)category— single category (optional)categories— multiple categories (optional)text— text search (optional)limit— max results (default 50)
Output: GeoJSON FeatureCollection + list items
get_location
Get detailed information for one location.
categories
List all supported search categories.
Supported Categories
- Restaurants, Coffee, Grocery, Parks, Museums
- Schools, Hospitals, Hotels, EV Chargers
Features
- ✅ Interactive MapLibre map with OpenStreetMap tiles
- ✅ Search by category, text, or both
- ✅ Adjustable search radius with visual overlay
- ✅ Pin markers that highlight on click
- ✅ Interactive results list (click to pan/zoom)
- ✅ Category filter chips
- ✅ Light and dark mode support
- ✅ Responsive layout (desktop + mobile)
- ✅ Full Zod validation on every request/response
- ✅ Strict TypeScript throughout
- ✅ Cloudflare Workers + Pages deployment
Deployment
See Deployment Guide for instructions on deploying to Cloudflare and registering with ChatGPT.
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。