hyperresearch
Hyperresearch is a deep research pipeline installed into Claude Code or OpenAI Codex as a skill or plugin. One prompt runs through a tier-adaptive 16-step pipeline and comes back as an adversarially audited report with full source provenance; bounded questions take a 5-step fast path, and an experimental dissertation mode writes 25K to 80K words. The README claims first place on the DRACO deep research benchmark, ahead of Claude Research, ChatGPT, Gemini and Perplexity, with method and per-task scores on its benchmark page.
Its checks are the main selling point. A cite-checker audits whether each cited source supports its sentence, and hallucinated quotes or unacknowledged retractions block the report. An independence audit collapses syndicated copies so reprints do not count as consensus. Four critics attack each draft and a tool-locked patcher can only make surgical edits. hpr scholar search queries OpenAlex, Crossref, CORE, DOAB, ClinicalTrials.gov, SEC EDGAR and FRED in one call, and for paywalled papers it looks for a legal open-access copy through Unpaywall, Europe PMC and CORE and discloses each substitution.
Every source lands in a markdown-plus-SQLite vault that later sessions search before fetching anything new, and crashed runs resume from their manifest. Install with pip install hyperresearch, then hyperresearch install for Claude Code or --target codex; Python 3.11 to 3.14 is required. A hosted version exists for people who do not want to run it locally. MIT licensed.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-26
- Commits, last 90 days
- 56
- Releases, last 90 days
- 4
- Latest release
- v0.12.0 (2026-09-24)
- Open issues
- 30
- Recent issues with at least one comment
- 2/9
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #31 Capture bibliographic metadata (journal, DOI, year, preprint-vs-published) for paper notes4 comments · 2026-06-08
- #59 Proposal: pluggable semantic index backend for larger vaults2 comments · 2026-07-24
- #34 refactor: extract lint rules into pluggable modules1 comments · 2026-06-17
- #178 Optional: publicly verifiable certificate for deep-research (one workflow file)0 comments · 2026-10-26
- #150 citecheck: numeric overlap auto-passes unsupported entity and metric claims0 comments · 2026-09-27
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Compatibility
Quick start
Use cases
Alternatives
Comments
No comments yet. Be the first!