Skip to main content
Alchemist is a research database that continuously ingests primary source documents, extracts structured insights from each one, and makes them retrievable by your agent. Every result carries the verbatim excerpt it was extracted from, the document it came from, the named author where one is credited, and the publication date. Your agent cites the primary source rather than a summary of a summary.

​
Core concepts

Insights

A single extracted finding from a document: a claim, a data point, or a directional statement. Each insight carries the exact quote it was derived from.

Documents

The primary source itself, whether that is a 10-K, a policy release, a data publication, or an analyst note. Each document has many insights, and you can retrieve a document with all of its insights in one call.

Sources

Who published it: an institution, a company, or an independent writer. Documents also carry their named authors where one is credited, so you can scope a search to a publisher or follow a particular author’s work.

Entity graph

Companies, agencies, people, and instruments named across the corpus, linked by the relationships the documents themselves state: suppliers, customers, competitors, regulators, and the events each entity appears in. Traverse from one entity to its neighbours to answer questions no single document contains.

​
Two ways to connect

​
What gets searched

Coverage spans corporate filings and disclosure, macro and policy analysis, energy and commodities, weather and climate data, and independent research from named analysts. Depth is uneven by vertical and changes as new sources are onboarded, so treat GET /api/documents/filters as the authoritative list rather than this page. It returns every source currently in the database.