Back to products
Perigon

Perigon

AI with real-time context, automated signals

Website perigon.io
Overview

What it is

Real-time news API and web content data — structured and enriched by AI, primed for LLMs.

Intent

I need it when

Identify major storylines and track how events evolve over time

Perigon's Stories endpoint clusters related articles by event or topic, auto-updates as stories evolve, and provides AI-generated summaries plus key points. This lets users pinpoint major narratives and measure coverage scope and reach without manually aggregating articles.

Filter news by editorial quality, bias, sentiment, and content type to reduce noise

Perigon tags every article with AI-powered labels (bias, opinion, paywalls, synthetic content), sentiment scores (positive/negative/neutral), topics, categories, and medium type. Users can include or exclude by any combination to surface only high-quality, relevant content and block sponsored or low-value pieces.

Monitor news and trends about specific companies, people, or topics in real time

Perigon's Articles endpoint ingests millions of news stories daily from 180,000+ global sources and allows filtering by company name, ticker, person name, Wikidata ID, or Boolean/vector search queries. Users can track exact mentions, exclude noise, and combine synonyms to catch all relevant coverage without manual monitoring.

Understand relationships between people, companies, and locations mentioned in news

Perigon's entity detection and knowledge graph mapping identifies real-world entities (people, companies, locations), resolves them across articles and sources, and traces relationships. Users can filter by entity to see how different actors connect across stories and time.

Build AI applications or conversational agents that answer natural-language questions about news

Perigon's vector search endpoint accepts plain-English prompts (e.g., 'Which AI startups raised Series A funding this week?'), returns semantically similar articles with relevance scores, and integrates via Model Context Protocol (MCP) into Claude, Cursor, or any MCP-compatible client. No complex Boolean syntax required.

Drop

Not a fit when

  • User needs historical news data older than 10 years for vector search (vector index limited to 6 months; older data requires /articles/all endpoint)
  • User requires video content as downloadable files or rich media streams (Perigon returns video results as article URLs only, not video feeds)
  • User needs real-time sentiment analysis for non-English languages outside English, French, German, Spanish, Portuguese, and Italian
  • User requires structured location data for non-US local publishers (sourceCountry filters currently return data for US-based local sources only)
  • User needs to filter by exact company or person without access to Wikidata IDs, journalist UUIDs, or canonical metadata lookups
Commercials

Pricing

Pricing not specified View pricing