wp-a/nature-academic-search

What it solves

Academic Paper Search provides a reproducible way to search, verify, and export academic literature. It solves the problem of "AI hallucinations" in citations by cross-referencing search results with real-world academic databases and performing field-level verification of metadata (titles, authors, years) to ensure citations are genuine and accurate.

How it works

The project implements a pipeline of Search $\rightarrow$ Deduplication $\rightarrow$ Verification $\rightarrow$ Export. It integrates with multiple academic sources including CrossRef, PubMed, arXiv, OpenAlex, and Europe PMC, with optional support for Semantic Scholar and ClinicalTrials.gov.

It is delivered as a set of tools that can be used in three ways:

1. MCP Server: Provides tools for AI clients (like Claude Code or Codex) to perform searches, resolve identifiers, and generate citations.
2. Skill: Guides AI models on how to decompose research questions into keywords and verify sources.
3. CLI: Allows for batch processing via YAML workflows, enabling automated retrieval and verification of large sets of papers.

Who it’s for

Researchers, academics, and AI power users who need to perform systematic literature reviews, verify the accuracy of AI-generated citations, or automate the collection of and export to reference managers like Zotero and EndNote.

Highlights

• Multi-source Retrieval: Simultaneously queries multiple databases to reduce gaps in literature discovery.
• Field-Level Verification: Compares expected citation data against actual database metadata to flag mismatches.
• Citation Graphing: Can trace upstream (references) and downstream (cited by) relationships to explore research networks.
• MeSH Integration: Supports lookup of Medical Subject Headings (MeSH) for precise biomedical searching.
• Automated Workflows: Supports YAML-defined pipelines for reproducible search and screening processes.