HyperResearch is a deep research system built around evidence, combining an open-source command-line tool on GitHub (about 3,900 stars) with a hosted research platform. It collects, searches and synthesizes web research into reports whose every citation is checked against its source, and keeps everything it reads in a persistent, searchable vault.
The open-source CLI plugs into coding agents such as Claude Code and Codex and turns them into deep research agents: a question is decomposed into chapters, sources are gathered from the open web plus more than 40 direct data APIs (sources like SEC EDGAR, PubMed, OpenAlex and official statistics), and findings are synthesized into a wiki that builds on itself across sessions. The hosted service runs the same stages at larger scale, adding managed fetching, a verification receipt for each report and access through a remote MCP server and REST API, so agents you already use can search the vault and start runs. The project is actively developed through 2026.
The GitHub project is free and open source under the MIT license for local use with your own agent. The hosted HyperResearch service is pay-as-you-go: Light runs cost $9 and Deep runs $49, with the first Deep run free, so the overall model is freemium.
HyperResearch is for analysts, researchers, investors and agent builders who need deep, well-sourced research reports and want the underlying evidence kept, searchable and verifiable over time.