GDELT Events MCP
guptaprakhariitr/gdelt-events-mcp
Detecção de eventos geopolíticos globais via GDELT 2.0 com análise de sentimento
Ver no GitHubServidor MCP que encapsula o GDELT 2.0 para detecção de eventos geopolíticos e macroeconômicos em tempo real. Fornece acesso a extração de eventos globais de notícias, análise de sentimento por timeline, tendências de atores e distribuição geográfica. Ideal para fundos macro, jornalismo de investigação e análise de risco geopolítico.
Instalação
npx convez-ai guptaprakhariitr/gdelt-events-mcp1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Análise de risco geopolítico para fundos macro e hedge funds
- Monitoramento de sentimento sobre empresas/países em notícias globais
- Rastreamento de tendências de atores e localizações de cobertura jornalística
- Análise ESG e de impacto social com distribuição geográfica de eventos
Veredito

Ferramenta sólida e bem-arquitetada para nicho específico (macro investing, jornalismo). GDELT é recurso gratuito subestimado; o valor real está na abstração de sua sintaxe complexa e na lógica de preprocessamento. Status de scaffold indica prototipagem avançada, não produção.
README
A tradução para o português deste README ainda está na fila.
gdelt-events-mcp — SCAFFOLD
Macro / geopolitical event detection. Wraps the GDELT 2.0 project — real-time event extraction from world news, with actor/location/sentiment metadata. Free underlying data; most people don't know GDELT exists.
Status: scaffolded. Idea #24 in ../../../ai-as-customer-ideas.md.
What GDELT is
- A free, ongoing project that ingests global news every 15 minutes, runs entity / event / sentiment extraction, and publishes the results as queryable APIs + bulk dumps.
- Schema includes: actors (Country, organization, person), event type (CAMEO codes), date, location, tone (-10 to +10), URL.
- Endpoint:
https://api.gdeltproject.org/api/v2/. - Bulk: 15-minute CSV files going back to 2015 (~500GB total, but daily slice is small).
Planned tools
| Tool | What it returns |
|---|---|
gdelt_search_events(query, country?, actor?, date_range, tone_range?) | Events matching filter; sorted by recency or relevance. |
gdelt_trending_actors(country, window?) | Most-mentioned actors in a country in last N hours. |
gdelt_tone_timeseries(query, granularity, range) | Sentiment timeseries for a topic (e.g. "How has news tone about trended?"). |
gdelt_geographic_heatmap(query, date_range) | Premium: geographic distribution of an event. |
gdelt_compare(query_a, query_b, range) | Premium: side-by-side volume + tone comparison. |
Audience
- Macro / hedge fund agents tracking geopolitical risk.
- Investor agents doing sentiment analysis on companies/countries.
- Newsroom agents finding "where is this story being covered?".
- ESG / impact agents tracking events in regions of interest.
Open / closed split
- Open: thin GDELT API wrapper, query builder.
- Closed: precomputed actor and topic indexes (the GDELT bulk data is overwhelming raw; the moat is preprocessing it).
Notes
- GDELT's own query syntax is quirky; abstracting it into a clean MCP surface is the value add.
- This product is niche but high-margin — small audience, but each customer pays well (macro funds, journalism teams).
- Easy to ship — GDELT API is REST/JSON, well-documented.
See also
../sec-edgar-mcp/— reference implementation; macro audience overlap.../README.md— Category 1 pipeline.
Sister MCPs
All from the same operator, all live on <product>.prakhar-cognizance.workers.dev, all free-tier friendly:
| Group | Products |
|---|---|
| Research | sec-edgar · arxiv · world-bank-economic · uspto-patents · fda-approvals |
| Verification + Utility | verification ⭐ · unit-converter |
| India | indic-normalize · indian-regulatory |
| Real-time | hn-trending · wikipedia-recent-changes · gdelt-events · crypto-prices |
| Healthcare | drug-interaction |
| Logistics | multi-carrier-tracking |
Full catalog: https://github.com/guptaprakhariitr · ⭐ = empty-quadrant / highest-conviction pick.