16-Step Research Pipeline for Claude Code

hyperresearch is a Python-based 16-step research pipeline that transforms Claude Code into a deep research agent. It focuses on solving the uncertainty of evidence verification and loss of research materials after investigation issues faced by existing AI research tools.

Key Features and How It Works

• 16-Step Precision Pipeline: The process is broken down into granular steps, from query decomposition to contradiction analysis, adversarial criticism, and citation verification (Cite-check).
• Strict Verification Gates: Prevents hallucinations by blocking publication if cited sentences do not exist in the source material or if retracted papers are cited.
• Revision-Centric Workflow: After report generation, instead of writing new drafts, it allows only surgical edits using the Read and Edit tools to enhance structural integrity.
• Sustainable Research Repository: All research materials are stored in SQLite and Markdown formats. These can be reused in subsequent investigations and accessed via MCP (Model Context Protocol) in Claude Desktop or Cursor.

Scale and Tone Adjustment Research scale can be adjusted to light, full, dissertation levels, and the tone can be set to teach, survey, analyze, or advocate, depending on the nature of the report.