worldmonitor: The Open-Source AI Global Intelligence Dashboard at 79K Stars

Anyone working in intelligence, investment research, risk control, or supply chain shares one pain: information is scattered across dozens of places. News on Reuters, geopolitical events in think-tank briefings, market data on Bloomberg, infrastructure status on status pages, social signals on X and Telegram. To assemble a picture of "what is happening in the world right now," you open twenty tabs, do the correlation in your head, and you are always half a beat late. worldmonitor wants to open-source and automate that process -- a real-time global intelligence dashboard that uses AI to pull 500+ news feeds, geopolitical signals, market data, and infrastructure status into one interface, running locally on Ollama with no API key required.

The repo koala73/worldmonitor was created on January 8, 2026. As of today (August 7, 2026) it sits at 79,487 stars, 11,881 forks, primary language TypeScript, AGPL v3 license, npm package name worldmonitor, with a push today -- seven months of unbroken iteration. That scale is near the front of the open-source dashboard category, and what is genuinely interesting is not the star count but that "unified situational awareness" has been seriously built out as an open-source project.

What It Is

worldmonitor (github.com/koala73/worldmonitor, docs at worldmonitor.app/docs/documentation) is positioned in one line: Real-time global intelligence dashboard -- AI-driven news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational-awareness interface.

It is not a news reader, nor a market terminal. It pulls the streams that are normally scattered -- news, geopolitical events, market data, infrastructure status -- into a single dashboard, then uses AI for cross-stream correlation and brief synthesis. It ships in two forms: a web app in the browser, and a Tauri 2 desktop app (macOS / Windows / Linux). The same codebase also spawns six site variants, covered below.

What Problem It Solves

The pain in one line: the inputs to global situational awareness are too scattered, and the correlation is too slow. Three layers:

1. Scattered sources. People working geopolitics, research, or supply-chain risk scan across news, official bulletins, market data, social media, and status pages every day. No single surface lays these streams side by side, let alone overlays them on one map.
2. Correlation done in the head. Military moves, economic data, disasters, and market anomalies have causal and lead-lag relationships, but traditional tools hand you raw feeds and leave the correlation to you -- slow, and easy to miss things.
3. Expensive or black-box. Commercial intelligence terminals (Bloomberg, Palantir, and the like) are expensive and closed; the data and processing logic are opaque. Building your own aggregation-plus-visualization-plus-AI-brief stack from scratch is real engineering work.

worldmonitor's entry point is an "open-source alternative" to the third item plus "aggregation plus AI correlation" for the first two. 500+ curated news feeds organized into 15 categories, with AI-synthesized briefs; cross-stream correlation converges four signal types -- military, economic, disaster, and escalation; a Country Instability Index layers on country-level pressure quantification. The whole system can run locally via Ollama with no external API key -- a hard requirement for sensitive scenarios where intelligence queries cannot go to the cloud.

Take a concrete scenario: a military move breaks out somewhere. The traditional flow is you see the news first, then tab over to market data to check the reaction, then to social media to gauge sentiment, then to a map to place it geographically -- four or five tools, and by the time you have correlated them, half an hour is gone. What worldmonitor does is lay those four streams side by side on the same map and the same timeline, with AI flagging the correlations, so you see one assembled picture instead of four scattered feeds.

Core Capabilities

The feature set from the README breaks into a few blocks:

A few designs worth calling out:

The dual map engine is not showmanship. The 3D globe suits global distribution and geopolitical layout; the WebGL flat map suits density and heatmaps. 56 layers means you can overlay news events, military moves, disasters, and financial data as needed -- this is the difference between "situational awareness" and "reading news."

CII v8 is server-authoritative. A country pressure score that each client computes independently would diverge across versions. worldmonitor makes it a server-authoritative value so everyone sees the same number. The 31 Tier-1 countries cover major economies and conflict hotspots.

Six site variants from one source is a smart engineering decision. world (general) / tech / finance / commodity / happy / energy -- six themes, one codebase with different configs, so each focus area gets a dedicated entry point without maintaining six codebases. The happy variant is notable -- a pure positive-news intelligence feed, an outlet for people who do not want to be drowned in bad news every day.

The finance radar is not just for reading quotes. 29 exchanges plus commodities and crypto, with a 7-signal market composite -- its positioning is not to replace Bloomberg for pricing, but to place market anomalies alongside other signals in an intelligence context. A country's pressure index rising while commodities move in tandem is exactly the kind of cross-domain correlation it wants to surface.

Cross-stream correlation is the core sell. Military, economic, disaster, and escalation signals each have no shortage of sources individually; the hard part is converging them in one view. A single event leaves different traces across different streams, and worldmonitor uses AI to stitch those traces together instead of making you flip through feeds one by one.

Architecture and Tech Stack

The tech choices show a serious TypeScript project:

• Map layer: globe.gl (3D globe, built on Three.js) + deck.gl (WebGL flat map, large-dataset visualization). These two are de facto standards in geo-visualization; together they cover both "looks good" and "handles volume."
• AI layer: Ollama for local inference, locked to no cloud model provider. Your intelligence queries do not leak to a third party. For risk-control and geopolitical scenarios, this is a privacy guarantee at the architecture level.
• Desktop shell: Tauri 2, not Electron. Tauri uses the system webview, with a smaller bundle and lower memory footprint and a Rust backend -- choosing Tauri over Electron signals the author cares about resource use; a dashboard app sits resident in the background, so this matters.
• Distribution: published as the npm package worldmonitor, with desktop installers also available.
• Internationalization: 26 languages + RTL, signaling a global audience rather than English-only -- an intelligence tool should have a global perspective by default.

Architecturally, the README stresses a "server-authoritative plus client-rendered" split: consistency-critical metrics like CII run server-side, while maps and visualization run client-side. The split is sound and prevents divergent data across clients. It carries a second meaning: the server-authoritative parts guarantee network-wide consistency, while the client-rendered parts can be customized and extended locally -- you can add layers and adjust styling in your own deployment without affecting metric consistency. For teams that want to build on it, this means "trustworthy data" and "editable presentation" can be had at the same time.

How to Get Started

The README's Quick start is three steps:

git clone https://github.com/koala73/worldmonitor
cd worldmonitor
npm install
npm run dev

For the local AI part, install Ollama and pull a model; worldmonitor discovers the local Ollama instance with no API key configuration. To run the desktop app, build via Tauri (macOS / Windows / Linux all supported); the exact command parameters are in the repo README and at worldmonitor.app/docs/documentation. The docs site covers configuration, layers, site variants, and local model integration -- worth a pass before you start.

Who It's For + Caveats

Who it's for

• People in geopolitics, investment research, risk control, supply chain, or emergency response who need a situational-awareness surface that aggregates multi-source signals in one place.
• Open-source intelligence (OSINT) practitioners: 500+ feeds plus cross-stream correlation plus map overlay is a ready-made OSINT workbench.
• Scenarios where queries cannot go to the cloud: runs on local Ollama, data stays on the machine.
• Anyone wanting to build on an existing dashboard: TypeScript + AGPL v3, with six site variants usable as templates.

Caveats

1. AGPL v3, not MIT / Apache. AGPL requires that modified versions served over a network also have their derivative source released. Using it for internal closed-source commercial work demands caution -- read the license terms first. This is its biggest difference from most starred open-source tools.
2. Local AI depends on Ollama. Local inference requires Ollama and a model installed, which is an on-ramp cost for anyone who has not touched local models. You can wire in an external model if you do not want the hassle, but then it is no longer "fully local."
3. Data sources are curated, not exhaustive. The 500+ feeds are selected -- broad coverage, but not the whole web. Niche verticals may need you to add sources yourself.
4. Server-side metrics like CII depend on the official service. The client-rendered parts run locally, but server-authoritative metrics (like CII v8) require the official service, so fully offline scenarios are limited there.

Take

Open-source dashboards with tens of thousands of stars are not rare; what is rare is "global intelligence situational awareness" -- something that used to belong to commercial terminals -- being seriously built out in the open. worldmonitor's value is not any single feature but the combination of "500+ source aggregation plus dual-map overlay plus cross-stream AI correlation plus local inference." Each piece exists elsewhere; the full set, open-source and locally runnable, is uncommon. AGPL v3 is a gate -- commercial teams must clear the license before adopting it. But for individual researchers, the OSINT community, and users in sensitive scenarios, this is a rare open-source intelligence workbench where you genuinely hold your own data and inference. Seven months of unbroken iteration, with a push today, says it is not a demo project.

There is also a point to be made about what it means for the "democratization of intelligence." Global situational awareness used to be a game only Bloomberg terminals, Palantir deployments, and government intelligence agencies could play; individual researchers and small organizations had to make do with manual stitching. worldmonitor opens that capability up, and lets it run locally -- it is not just about saving money, it is about being able to audit the data sources and the correlation logic, which in intelligence work matters more than the features themselves. The choice of AGPL v3 fits this positioning: it prevents anyone from taking the open-source work, closing it, and reselling it, protecting the public character of open-source intelligence.

References

• worldmonitor GitHub repo: https://github.com/koala73/worldmonitor
• worldmonitor documentation: https://worldmonitor.app/docs/documentation
• Star / fork / language / license / created time / most recent push per GitHub API (verified 2026-08-07, 79,487 stars)
• Features, map engines, CII, site variants, local AI per the repo README (2026-08-07)