Hyperresearch

Hyperresearch is an open-source deep research harness for Claude Code. One prompt routes through a tier-adaptive pipeline — light path for bounded questions, full path with parallel drafts and four critics, optional dissertation scale for chapter-length work. It pulls from eight scholarly surfaces in one client, prefers full open-access text over abstract-only paywalls when legal copies exist, and blocks hallucinated quotes at the gate. Crashed runs resume from a manifest.

Founders still paste a vague ask into chat and get a pretty report with soft citations. Hyperresearch treats research like a product: width sweep, contradiction graphs, adversarial critics, and a hard cite-check before anything ships. Sources land in a searchable vault so the next session starts smarter instead of re-fetching the same web.

In a project folder run pip install hyperresearch && hyperresearch install, then use /hyperresearch with your question inside Claude Code. The router classifies the tier, loads one step skill at a time, writes notes and reports into the vault, and lets you resume if a long run dies mid-pipeline. Power users can install globally so the skill is available in every session.

Not another coding TUI and not a one-shot web-search skill. The product is the audited research loop plus a durable source vault — independence audits so syndicated reprints do not fake consensus, tool-locked patchers that cannot rewrite the whole draft, and a pipeline built to survive long runs without dropping steps as context rots.

• Agent Reach
• OpenWiki
• Spec Kit