APITube turns the world's news into structured JSON. It aggregates articles from 500,000+ verified sources across 180+ countries in 60+ languages, and enriches every article before you ever see it: sentiment scores, extracted entities (people, organizations, locations, brands, events), IPTC categories, topics and industries. You call one HTTP endpoint and get clean data back — no crawler to maintain, no HTML to parse, no NLP pipeline to train.
What you actually call
Search runs through GET /v1/news/everything with 65+ filter parameters that combine in a single request: keywords, language, country, source domain, source quality rank, sentiment range, entity IDs, media presence and date ranges. Other endpoints cover the shapes people ask for most often:
/v1/news/top-headlines — breaking coverage by country, language or category
/v1/news/story — articles already clustered into one story, so you get the event instead of forty near-duplicates
/v1/news/trends — trending entities, topics and categories over time buckets
/v1/news/stream — a live Server-Sent Events feed that pushes matches as they are indexed
/v1/suggest/entities — autocomplete an entity ID before you filter on it, so "Tesla" becomes a stable ID instead of a fragile keyword
Authentication is a single API key, passed as an api_key query parameter, an X-API-Key header, or an Authorization: Bearer header. Responses are JSON by default; paid plans also export XML, CSV, TSV, XLSX, JSONL and Parquet. Webhooks and WebSocket dispatches push new matches to your own endpoint as they arrive.
Built for AI agents too
There is a hosted MCP server at https://mcp.apitube.io/ — nothing to install, no local process to babysit. Point Claude, Cursor, VS Code or any other MCP client at it with an API key in the Authorization header and the model gets two tools: search_news, which exposes every News API filter, and suggest, which turns a name like "Tesla" into the entity ID those filters expect. Four prompts ship with the server: monitor_company, topic_sentiment, breaking_news and compare_coverage. The MCP server requires a paid plan, starting at Starter.
Who it is for
Brand and media monitoring — track mentions of a company or product across languages, with sentiment already attached
Market and financial intelligence — filter by entity, industry and sentiment to feed risk or trading models
RAG pipelines and AI agents — give an LLM a live, structured news corpus instead of scraped HTML
Research and data science — pull historical coverage for backtesting and longitudinal analysis
Products with a news feed — apps, dashboards and newsletters without running a crawler yourself
Pricing
The free tier gives 100 requests per day and needs no card. Paid plans start at $29/month (Starter, 10,000 requests/month), then $99 (Basic, 50,000) and $199 (Professional, 150,000). Annual billing is 20% off, and pay-as-you-go is $0.01 per request. SDKs for Node.js, PHP and Python live at github.com/apitube, and the full parameter reference is at docs.apitube.io.
