BGPT Paper Search

Problem

Literature search tools often return only titles, abstracts, and a small set of metadata. Systematic reviews, scoping reviews, and evidence tables usually need more specific fields: methods, quantitative results, sample sizes, effect sizes, and quality scores. When those details are buried in full text, manual extraction becomes slow, and cross-study comparison is harder because terminology and reporting are inconsistent.

How It Works

BGPT Paper Search connects to a curated scientific paper database built from full-text raw experimental data through a remote MCP service, so there is no local paper repository to install. The skill calls the search_papers tool and returns structured results, typically including:
- Basic metadata: title, authors, journal, year, DOI
- Methods: experimental technique, model, or protocol
- Results: key findings with quantitative data
- Sample size: number of subjects or samples
- Quality score: research quality assessment
- Conclusion: author conclusions and implications

This output is better suited for evidence tables, comparison matrices, or synthesis workflows than manually inferring fields from abstracts.

Boundaries and Notes

It is best for deep retrieval where experimental detail, quantitative outcomes, or quality grading matters. For very broad initial discovery or the latest preprints, it may be worth pairing with other sources. The service depends on network access and remote MCP availability; the free tier has a search limit, and paid results require an API key. search_papers returns structured fields from the BGPT database, not arbitrary direct PDF scraping.