BGPT - Scientific Paper Search

This connects Claude or any MCP client to a database of full-text scientific papers, exposing a search_papers tool that returns structured experimental data instead of just abstracts. You get 25+ fields per paper: methods, results, conclusions, quality scores, sample sizes, and limitations. It runs as a remote service over SSE or streamable HTTP, so there's no local setup. The same API is available as a plain REST endpoint if you're calling from Python notebooks or scripts. Free tier gives you 50 results, then it's pay-as-you-go at two cents per result. Useful when you need your AI to reason over actual experimental findings rather than surface-level metadata, or when you're building literature reviews and evidence synthesis workflows.