worldmonitor-mcp
MCP server wrapping worldmonitor global intelligence dashboard, exposing 140 tools across 32 services for live market, geopolitical, military, cyber, climate, and supply chain data.
README
worldmonitor-mcp
MCP server and CLI that wraps koala73/worldmonitor, an open-source global intelligence dashboard. This exposes 140 tools across 32 services to Claude Code, Cursor, Windsurf, and any other MCP client so you can ask models for live market, geopolitical, military, cyber, climate, and supply chain data.
Most services proxy the public worldmonitor.app API. A handful bypass it and hit underlying data providers directly (SEC EDGAR, FRED, Finnhub, USDA, NOAA, CFTC, and others). Those are the extensions. The dashboard itself isn't mine, koala73 built that. This repo is just the MCP wrapper plus a few direct additions.
Install
# as a one-off
npx worldmonitor-mcp
# or install globally
npm install -g worldmonitor-mcp
To use from Claude Code, add this to your MCP config.
{
"mcpServers": {
"worldmonitor": {
"command": "npx",
"args": ["worldmonitor-mcp"]
}
}
}
The CLI also works as a standalone tool.
worldmonitor list-earthquakes
worldmonitor list-market-quotes --symbols AAPL,NVDA
worldmonitor get-fred-series --series GDP
What's in it
32 services grouped by what they cover.
| Category | Services |
|---|---|
| Markets and financial | market, economic, treasury, cftc, congress, onchain, sentiment |
| Geopolitics and intel | intelligence, conflict, military, unrest, displacement |
| News and research | news, research, article |
| Environmental | climate, weather-agriculture, wildfire, seismology |
| Supply chain and trade | supply-chain, maritime, trade |
| Infrastructure and cyber | infrastructure, cyber, aviation |
| Government and regulation | government, sec-edgar, economic-calendar |
| Humanitarian | giving, positive-events |
| Predictions | prediction |
| Legacy bulk and cache | legacy |
21 of the 32 services proxy the upstream WorldMonitor API. 10 services hit underlying providers directly (SEC EDGAR, FRED, Finnhub, USDA, NOAA, CFTC, CBOE, Treasury Fiscal Data, Fear & Greed, Federal Register, Reddit, Gov Contracts, OpenSanctions). The legacy service mixes Vercel serverless and a Railway relay for rate-limited upstream APIs.
Env vars
None of these are required. Defaults work out of the box.
| Variable | Purpose | Default |
|---|---|---|
WORLDMONITOR_BASE_URL |
Upstream worldmonitor.app URL | https://worldmonitor.app |
WORLDMONITOR_API_KEY |
Upstream API key if you have one | none |
WORLDMONITOR_TIMEOUT |
HTTP timeout in ms | 30000 |
WORLDMONITOR_MAX_RESPONSE_SIZE |
MCP response size soft cap in bytes | 100000 |
FINNHUB_API_KEY |
Earnings/IPO calendar, market quotes | none (optional) |
FRED_API_KEY |
FRED economic data | none (optional) |
USDA_API_KEY |
Crop and drought reports | none (optional) |
OPENSANCTIONS_API_KEY |
Sanctions search enhancement | none (optional) |
The direct-API services mostly work without any keys. Keys unlock the rate-limited or paid endpoints.
How it works
The wrapper reads a service registry (src/services/*) where each service declares its tools, endpoints, and parameter schemas. At startup the MCP server iterates every service and registers every tool. When a tool is called, routing splits two ways. Direct handlers bypass the upstream and call the underlying data provider. Proxy handlers forward to the configured WorldMonitor base URL.
On top of that there's retry-with-backoff on 429 and 5xx responses, response validation that catches broken serverless endpoints returning HTML instead of JSON, known-broken endpoint detection that returns a helpful error instead of a confusing failure, and a response truncation layer that shrinks oversize arrays to fit the MCP context window.
Article extraction has three fallbacks. Direct fetch first, then Google Cache, then Archive.org, with a browser user-agent to bypass soft paywalls.
Credits
The upstream dashboard is koala73/worldmonitor. It's a real-time global intelligence dashboard with 100+ news feeds, a 3D deck.gl WebGL globe, AI briefs with local LLM support, and Tauri desktop builds for macOS, Windows, and Linux. Licensed AGPL-3.0. If you find this MCP useful, that's the project doing the heavy lifting.
This repo is just the MCP wrapper and the direct-API extensions. No data collection or aggregation lives here.
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。