Hi! This project is really interesting — evaluating MCPs for scientific literature retrieval is exactly the kind of benchmarking the ecosystem needs.
You might want to include BGPT MCP in the evaluation. It takes a different approach from the PubMed/arXiv MCPs — instead of returning abstracts, it extracts structured experimental data from full-text papers: methods, results, sample sizes, quality scores, conclusions, and 25+ other fields per paper.
Config:
{
"mcpServers": {
"bgpt": {
"url": "https://bgpt.pro/mcp/sse"
}
}
}50 free searches, no API key needed. Would be interesting to see how structured data extraction compares to abstract-based retrieval in your evaluation framework.
Hi! This project is really interesting — evaluating MCPs for scientific literature retrieval is exactly the kind of benchmarking the ecosystem needs.
You might want to include BGPT MCP in the evaluation. It takes a different approach from the PubMed/arXiv MCPs — instead of returning abstracts, it extracts structured experimental data from full-text papers: methods, results, sample sizes, quality scores, conclusions, and 25+ other fields per paper.
Config:
{ "mcpServers": { "bgpt": { "url": "https://bgpt.pro/mcp/sse" } } }50 free searches, no API key needed. Would be interesting to see how structured data extraction compares to abstract-based retrieval in your evaluation framework.