BGPT MCP
Search scientific papers built from full-text experimental data via hosted MCP server. 50 free searches, no API key needed. [Free]
Last verified:
What is BGPT MCP?
BGPT MCP is a remote Model Context Protocol (MCP) server that gives AI assistants access to a comprehensive database of scientific papers built from full-text studies. Unlike traditional search engines that return only titles and abstracts, BGPT extracts raw experimental data from the full text of published papers—including actual data, methods, results, quality scores, sample sizes, and 25+ other metadata fields. This enables AI tools like Claude Desktop, Cursor, and Claude Code to reason over real scientific findings grounded in evidence rather than summaries.
The tool provides a single feature called search_papers using SSE transport, allowing users to search for papers by query terms, filter by recency with days_back, and look up specific papers by DOI. Each search result returns over 25 organized fields covering identification (DOI, title, publication date, journal), content and analysis (one-sentence summary, keywords, results and conclusions, methods and experimental techniques, sample size and population, extracted data list), and critical evaluation (how to falsify, study blindspots, limitations and biases, conflict of interest). Additional fields include technical details like experimental models, software and tools used, data availability statements, code and data links, lab name, funding information, citations, reference count, and quality scores for scientific merit, novelty, generality, usefulness, and reproducibility.
BGPT MCP is designed for researchers, scientists, graduate students, biotech professionals, and anyone using AI assistants for evidence synthesis, literature reviews, or research assistance. It integrates seamlessly into existing AI workflows by adding a simple configuration to Claude Desktop, Cursor, or any MCP-compatible client. The founder created BGPT post-Berkeley to bridge computer science and immunology, ensuring every output is verifiable and bias-aware with epistemic humility built into every answer.
The tool forces skepticism at its core by synthesizing from PubMed, patents, and bio-banks, with a
BGPT MCP pricing
Pricing model: Freemium
Free Tier: $0 for 50 free results with all paper metadata fields and no API key needed. Pay-as-you-go: $0.02 per result with unlimited searches, all paper metadata fields, and metered Stripe billing. After exhausting the 50 free results, users must subscribe through Stripe for continued access.
BGPT MCP pros
- 50 free searches with no API key required
- Extracts raw experimental data from full-text papers, not just abstracts
- Returns 25+ organized metadata fields per paper
- Includes quality scores for scientific merit, novelty, generality, usefulness, and reproducibility
- Integrates with Claude Desktop, Cursor, Claude Code, and any MCP-compatible tool
- No local installation needed—hosted remote MCP server
- No sign-up required for free tier
- Pay-as-you-go pricing at $0.02 per result
- Look up specific papers by DOI directly
- Filter searches by recency using days_back parameter
- Includes critical evaluation fields like limitations, biases, and conflicts of interest
- Provides how_to_falsify and study_blindspots for epistemic humility
- Daily database updates with new scientific literature
- Metered Stripe billing for pay-as-you-go plans
- Includes funding, citations, and reference count data per paper
BGPT MCP cons
- Only 50 free results before payment required
- $0.02 per result can add up quickly for extensive research
- Requires configuring MCP server in AI tool manually
- No web interface—must use through MCP-compatible AI clients
- Pay-as-you-go model may be expensive for heavy users
- Limited to scientific papers in biology and related fields
- No free unlimited plan available
- API key management needed for paid tier
Frequently asked questions about BGPT MCP
How do I start searching for free?
Just add the MCP configuration to your AI tool and start asking questions. No sign-up and no API key is needed. You automatically get 50 free results when you begin using BGPT MCP.
What happens when I use up my 50 free results?
After using your 50 free results, you need to subscribe through Stripe for pay-as-you-go access at $0.02 per result returned. Billing is metered through Stripe.
How do I subscribe and get my API key?
Subscribe through the Stripe subscription link provided on the BGPT website. After subscribing, you will receive an API key which is your Stripe subscription ID used for paid access.
How do I use my API key?
Include your API key (Stripe subscription ID) in the optional api_key parameter when calling the search_papers tool. This enables paid searches after your free tier is exhausted.
How does billing work exactly?
Billing is metered through Stripe at $0.02 per result returned. You only pay for the results you actually receive, and billing is handled automatically through your Stripe subscription.
What does days_back do?
The days_back parameter is an optional filter that limits search results to papers published within the specified number of days. This helps you find recent research in your field.
Where do I find my API key if I lost it?
Your API key is your Stripe subscription ID. You can find it in your Stripe account dashboard under subscriptions, or contact BGPT support for assistance recovering it.
Can I cancel anytime?
Yes, you can cancel your Stripe subscription at any time. After cancellation, you will revert to the free tier with 50 results if you have unused free searches available.
What transport does BGPT MCP use?
BGPT MCP uses SSE (Server-Sent Events) transport at https://bgpt.pro/mcp/sse. It also provides a Streamable HTTP endpoint at https://bgpt.pro/mcp/stream for compatible clients.
What fields are included in each paper result?
Each result includes 25+ fields: DOI, title, publication date, journal, one-sentence summary, keywords, results and conclusions, methods and experimental techniques, sample size and population, extracted data list, problem statement, study context, how_to_falsify, study_blindspots, limitations and biases, conflict of interest, taxonomy information, experimental models, software and tools used, data availability statements, code and data links, lab name, funding JSON, citations JSON, reference count, and quality scores for scientific merit, novelty, generality, usefulness, and reproducibility.