Scry - Query 600 GB indexes over Hacker News, ArXiv, and more using Claude Code.

What it does

Scry is a queryable index service that lets users run SQL queries over large public datasets—including Hacker News, ArXiv, Reddit, PubMed, Wikipedia, and many more—via a SQL-over-HTTPS API. It is designed to be used by AI agents (such as Claude Code) to retrieve structured, source-attributed evidence from over 50 corpora totaling hundreds of gigabytes. Queries support full-text search (BM25), vector embeddings, hybrid search, temporal deltas, and concept algebra. Results include source, author, timestamp, and URL for verifiability.

Who it is for

Scry is built for researchers, analysts, and AI developers who need to systematically search and synthesize public evidence. It is especially suited for users who want to give AI agents direct, inspectable access to large text corpora rather than relying on ad hoc web search. The platform is in alpha and requires an API key; usage is metered via credits.

Why it matters

Traditional web search gives agents a handful of sampled links and opaque summaries. Scry provides explicit source scope, composable SQL queries, and durable execution receipts. This enables reproducible, verifiable research workflows—for example, tracking how a term spread across forums over time, or finding authors who praised one model release and criticized the next. The project positions itself as a public-benefit epistemic infrastructure for transparent evidence retrieval.

Launch signal

Scry was announced on Hacker News as a Show HN on exopriors.com/scry. The landing page describes it as "premium infrastructure for serious research work" and lists a large corpus table with sources, record counts, and freshness indicators. The project is in alpha, with beta notifications available.

Brand and naming

The name "Scry" evokes the idea of seeing hidden or distant information, fitting for a tool that surfaces structured evidence from vast archives. The brand emphasizes transparency, composability, and agent-native access. The tagline "Query 600 GB Indexes over Hacker News, ArXiv, etc. using Claude Code" clearly positions it as a developer tool for AI agents.