This is an MCP server that searches scientific papers and returns structured experimental data extracted from full text, not just abstracts. You get 25+ fields per paper including methods, sample sizes, quantitative results, and quality scores. It's useful for lit reviews when you need the actual numbers and protocols buried in papers, or when building evidence tables for meta-analyses. Runs remotely via npx, no local install needed. Free tier gives you 50 searches per network, then it's a penny per result. The fact that it pulls from full-text extraction rather than just metadata makes it more useful than hitting PubMed, assuming the database covers your field.
npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill bgpt-paper-search