TaewoooPark/scholar-megasearch

View on GitHub ↗

Massive multi-source academic literature search for Claude Code — one skill fans out subagents across 20+ scholarly databases (arXiv, Semantic Scholar, Crossref, OpenAlex, PubMed, …), merges into a deduplicated ranked corpus, and acquires the original PDFs.

24 ★8 forksPythonUpdated 21d ago

What you need to know

A skill + MCP setup for massive multi-source academic literature search in Claude Code and Codex. Fans out subagents across 20+ scholarly databases, merges into a deduplicated, corroboration-ranked corpus, and downloads original PDFs.

Install

git clone https://github.com/TaewoooPark/scholar-megasearch.git
bash setup/install.sh you@example.com  (or --target codex / --target both)

Usage

  • Trigger the skill in natural language, e.g. "search every database for graph neural networks and grab the PDFs"
  • Or use /scholar-megasearch depth=4 <topic> as a slash command

Key features

  • 20+ sources in one pass with parallel subagent fan-out
  • Dedup by DOI/arXiv/title with provenance tracking
  • Corroboration ranking and optional --min-sources 2 shortlist
  • Automatic PDF acquisition via open-access routes
  • Depth levels L1-L5 with recursive citation snowballing

Best for

Systematic, broad literature sweeps with PDFs from Claude Code/Codex

Caveats

  • Requires Claude Code or Codex and Python 3.11+ with uv
  • paper-search-mcp must be the git-main build; Asta rate limits apply
  • PDF acquisition is open-access-first; paywalled papers are flagged
Platforms: macOS · LinuxClients: Claude Code · Codex

Reviewed 2026-08-11

Topics

Stars
24★
Forks
8
Language
Python
License
MIT
Created
2026-05-29
Last push
2026-08-14