JSON Recon

Overview

JSON Recon is an AI-powered schema.org JSON-LD extraction API for autonomous AI agents. Extract structured data from any URL as clean, validated JSON-LD — no scraping, no parsing, no markdown cleanup. Supports 200+ schema.org types including Product, Restaurant, Event, LocalBusiness, Article, Recipe, JobPosting, MedicalCondition, SoftwareApplication, and more. Returns specification-compliant structured data with confidence scoring, source provenance, and automatic type detection. Built for agent-to-agent commerce via x402 on Base (USDC micropayments). Three endpoints: /extract for full AI extraction (.01), /extract/lite for lightweight JSON-LD parsing (.001), and /extract/batch for multi-URL batch processing (.05). Specialized domain expertise at zero LLM cost: Reddit via native JSON API for posts, comments, subreddits, and profiles (QAPage, DiscussionForumPosting, CollectionPage). Google Patents with patent metadata, inventors, assignees, citations, and PDF links. Academic coverage across 30+ publishers including PubMed, Nature, arXiv, IEEE, Springer, PLOS, Wiley, ScienceDirect, SAGE, Frontiers, MDPI, bioRxiv, medRxiv, SSRN, JSTOR, and ACM for ScholarlyArticle with authors, DOI, journal hierarchy, and references. Four-tier scraping architecture: HTTP pre-check for instant JSON-LD detection, headless browser for JavaScript rendering, stealth browser for anti-bot protected sites like Amazon and Zillow, plus HTTP fallback for maximum coverage. Tiered extraction guarantees a response: existing JSON-LD first, then AI-powered extraction via Gemini, then metadata fallback. Redis caching for sub-10ms repeats. Ideal for web research, competitive intelligence, price monitoring, content aggregation, academic research, patent analysis, SEO auditing, knowledge graphs, and real-time data enrichment. Token-efficient alternative to raw HTML scraping and markdown conversion.

Health