global-sentinel-mcp
Unifies prediction markets, global news events, aviation state, and internet-infrastructure telemetry into a single async MCP server for macro and financial analysis.
README
global-sentinel-mcp
<!-- mcp-name: io.github.ykshah1309/global-sentinel-mcp -->
An async MCP server unifying prediction markets, global news events, aviation state, and internet-infrastructure telemetry into one alternative-data layer for macro and financial analysis.
Status: Beta — ready for production use; public API may still shift before 1.0.
Why this exists
Alternative data — prediction-market odds, real-time geopolitical event streams, live aviation transponder data, and internet-infrastructure telemetry — is increasingly central to macro and financial analysis. Each source has its own API shape, authentication model, and update cadence, making it tedious to wire them together for every new project.
global-sentinel-mcp closes that gap by wrapping four public data feeds behind
a single Model Context Protocol (MCP) interface. An LLM agent can ask
probability, event, aviation-state, and internet-telemetry questions in one
place, without knowing the upstream API details. It also exposes a composite
macro alert tool that fuses news events and outage annotations into one
convergence signal with a human-readable explanation.
The server is fully async (httpx.AsyncClient with retry transport), returns
typed Pydantic v2 models for every response and error, ships a TTL cache per
source to stay within upstream rate limits, and installs in one command over
stdio — no web server required.
Available Tools
| Tool | Data Source | What it returns |
|---|---|---|
get_prediction_odds |
Polymarket | Ranked event odds (probability %, volume, end date) for a query |
get_prediction_event_by_id |
Polymarket | Odds for a specific event id or slug |
lookup_aircraft_state |
OpenSky Network | Latest ADS-B state for a tail number (N-Number) or ICAO24 hex |
convert_nnumber |
— (offline) | Pure-Python FAA N-Number ↔ ICAO24 conversion, no network needed |
get_global_events |
GDELT 2.0 | High-impact events for a country, ranked by Goldstein scale |
check_network_anomalies |
Cloudflare Radar | BGP hijacks + leaks and regional outage annotations |
get_macro_alert |
GDELT + Cloudflare | Composite convergence signal with explanation string |
Quickstart
1. Install
uvx global-sentinel-mcp
Or install from PyPI:
pip install global-sentinel-mcp
2. Set environment variables (all optional — unset tools degrade gracefully)
# OpenSky — OAuth2 preferred (post-2025 migration), Basic Auth for legacy accounts.
export OPENSKY_CLIENT_ID="your_opensky_client_id"
export OPENSKY_CLIENT_SECRET="your_opensky_client_secret"
# Cloudflare Radar — required for BGP and outage tools.
export CLOUDFLARE_API_TOKEN="your_cf_token"
# Optional.
export LOG_LEVEL="INFO"
3. Run
global-sentinel-mcp
The server speaks MCP JSON-RPC over stdio.
Local development
git clone https://github.com/ykshah1309/global-sentinel-mcp
cd global-sentinel-mcp
uv sync --dev
uv run global-sentinel-mcp # run the server
uv run pytest tests/ -v # run the test suite
uv run ruff check src/ tests/ # lint
uv run mypy src/ # type-check
Claude Desktop configuration
Add to claude_desktop_config.json (see claude_desktop_config.example.json):
{
"mcpServers": {
"global-sentinel": {
"command": "uvx",
"args": ["global-sentinel-mcp"],
"env": {
"OPENSKY_CLIENT_ID": "your_opensky_client_id",
"OPENSKY_CLIENT_SECRET": "your_opensky_client_secret",
"CLOUDFLARE_API_TOKEN": "your_cf_token"
}
}
}
}
Cursor configuration
Add to .cursor/mcp.json:
{
"mcpServers": {
"global-sentinel": {
"command": "uvx",
"args": ["global-sentinel-mcp"],
"env": {
"OPENSKY_CLIENT_ID": "your_opensky_client_id",
"OPENSKY_CLIENT_SECRET": "your_opensky_client_secret",
"CLOUDFLARE_API_TOKEN": "your_cf_token"
}
}
}
}
Environment Variables
| Variable | Required for | Description |
|---|---|---|
OPENSKY_CLIENT_ID |
aviation (OAuth2, preferred) | OpenSky OAuth2 client id — register free |
OPENSKY_CLIENT_SECRET |
aviation (OAuth2, preferred) | OpenSky OAuth2 client secret |
OPENSKY_USERNAME |
aviation (legacy) | OpenSky Basic Auth username (pre-2025 accounts) |
OPENSKY_PASSWORD |
aviation (legacy) | OpenSky Basic Auth password |
CLOUDFLARE_API_TOKEN |
BGP + outages | Cloudflare API token with Radar read permissions |
LOG_LEVEL |
— | Python log level (DEBUG / INFO / WARNING / ERROR). Default INFO. |
Polymarket and GDELT tools require no authentication. Anonymous OpenSky is allowed but heavily rate-limited.
Data Sources
Polymarket (Gamma API)
Crowd-sourced prediction market data. Token prices reflect the crowd's
estimated probability of future events. No API key required. Events expose
outcomes and outcomePrices as JSON-encoded strings — the client parses the
"Yes" leg and returns it as probability_pct.
Docs
GDELT 2.0
The world's largest open-access global news event database, updated every 15 minutes. Events are coded with actors, event types, Goldstein conflict/cooperation scores, and geographic coordinates. Per publish cycle the server downloads and parses the 5 MB export zip exactly once, then answers all country queries from an in-memory DataFrame cache. Docs
OpenSky Network
Public ADS-B flight tracking from a global network of volunteer receivers. Returns real-time position, altitude, velocity, and on-ground status for any aircraft with a Mode S transponder. Post-2025 the service moved to OAuth2 client-credentials; this server handles token refresh automatically and falls back to Basic Auth (legacy) or anonymous if credentials are absent. Docs
Cloudflare Radar
Internet-infrastructure telemetry from Cloudflare's global network. Covers BGP
routing anomalies (/bgp/hijacks/events, /bgp/leaks/events) and regional
outage annotations (/annotations/outages, formerly "traffic anomalies",
renamed by Cloudflare in 2024). Requires a free Cloudflare API token.
Docs
Contributing
See CONTRIBUTING.md for setup instructions, branch naming conventions, testing requirements, and code style guidelines.
License
MIT — Copyright (c) 2026 ykshah1309
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。