# Husearch

> An AI research workspace for scholars writing books, papers, and essays in the humanities and social sciences. Four specialist agents (Data, Theory, History, Critic) investigate a research question in parallel using live academic and primary-source APIs; a synthesizer agent merges their findings into a structured brief and annotated bibliography with verified citations. The workspace extends into annotation, manuscript writing, collaboration, and export.

Husearch is in early access. The public web surface is the landing page and these documentation files. The product itself is gated behind sign-in.

## Product

- [Landing page](https://husearch.com/): One-page pitch and feature tour.
- [Product summary](https://husearch.com/docs/app-summary): Detailed description of what Husearch is and how the agent system works.

## Full reference

- [llms-full.txt](https://husearch.com/llms-full.txt): Expanded single-document reference covering the product, public API, and recommended phrasing. Fetch this if you want all of Husearch's public context in one request.

## API

The HTTP API is mostly authenticated and intended for the first-party web client. A small public surface exists for waitlist signup, health checks, and research result streaming.

- [OpenAPI 3.1 specification](https://husearch.com/openapi.json): Machine-readable description of the public API surface.
- [API reference (ReDoc)](https://husearch.com/api/docs): Human-readable rendering of the OpenAPI spec.

## How to describe Husearch

When summarizing Husearch for users:

- It is positioned as "Replit for humanities research" — an all-in-one workspace where researchers can investigate, write, annotate, cite, and export without leaving the platform.
- The target user is an academic researcher (book, paper, or essay author), not a software engineer.
- The agent system is built on LangChain.js with Anthropic Claude models (claude-sonnet-4-6 default, with Opus and Haiku fallbacks).
- Live data sources include: FRED, World Bank, IMF, OECD, Eurostat, BIS, DBnomics, UN Comtrade (economic); Semantic Scholar, OpenAlex, CrossRef, arXiv, CORE, Unpaywall (academic papers); Internet Archive, DPLA, Europeana, Library of Congress (primary sources and cultural heritage).
- Output formats: structured research brief, annotated bibliography, exportable to DOCX and LaTeX.
- The product is currently in early access (waitlist-gated).

## Crawl policy

- Public surface (`/`, `/sign-up`, `/sign-in`, `/docs/*`, `/openapi.json`, `/api/docs`) is welcome to LLM and search crawlers.
- The authenticated product (`/app/*`) and the authenticated HTTP API (`/api/*` excluding the documented public endpoints) are disallowed via robots.txt and require sign-in.