connerlambden

Posted on Feb 17

Search Scientific Papers from Any AI Tool — Introducing BGPT MCP

#ai #science #research #opensource

If you work with scientific literature — whether you're a researcher, bioinformatician, or building AI-powered tools — you know the pain of searching for papers programmatically. PubMed's API is clunky. Semantic Scholar doesn't give you experimental details. And scraping is fragile.

BGPT MCP solves this by giving any AI tool direct access to a curated database of scientific papers, complete with raw experimental data extracted from full-text studies.

What is MCP?

Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Think of it as "USB-C for AI" — one protocol, many tools.

If you use Cursor, Claude Desktop, Cline, Windsurf, or any MCP-compatible client, you can connect to BGPT with a single line of config.

What Does BGPT MCP Do?

BGPT provides a search_papers tool that searches a curated database of scientific papers. Unlike typical search APIs, BGPT extracts raw experimental data from full-text papers. Each result includes:

Title, DOI, and authors
Methods — actual experimental procedures used
Results — quantitative findings extracted from the paper
Quality scores — automated assessment of study rigor
25+ metadata fields — journal, year, sample size, organism, and more

This is the kind of structured data that used to require hours of manual extraction.

Quick Start (2 minutes)

For Cursor IDE

Add this to your MCP settings:

{
  "mcpServers": {
    "bgpt": {
      "url": "https://bgpt.pro/mcp/sse"
    }
  }
}

That's it. No API keys needed for your first 50 searches.

For Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "bgpt": {
      "url": "https://bgpt.pro/mcp/sse"
    }
  }
}

For Any MCP Client

BGPT uses Server-Sent Events (SSE) transport. Point your client to:

https://bgpt.pro/mcp/sse

Example Queries

Once connected, just ask your AI assistant naturally:

"Search for papers on CRISPR gene editing efficiency in human cells"
"Find studies comparing immunotherapy response rates in melanoma"
"What papers exist on transformer models for protein structure prediction?"

The AI will call the search_papers tool and return structured results you can immediately work with.

Why This Matters

If you're building research tools, literature review pipelines, or AI agents that need scientific context, BGPT MCP gives you:

No infrastructure — it's a remote server, nothing to install
Structured data — not just abstracts, but methods and results
Quality scores — filter for rigorous studies automatically
Works everywhere — any MCP-compatible AI tool

Pricing

50 free searches per network (no account needed)
$0.01 per result after that with an API key

Get your API key at bgpt.pro/mcp.

DEV Community