playbooks

home / mcp / research mcp server

Research MCP Server

Provides MCP-based access to arXiv, Semantic Scholar, and PubMed for live search, deduplication, citation analysis, and BibTeX generation.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "gyash1512-researchmcp": {
      "command": "npx",
      "args": [
        "-y",
        "researchmcp",
        "arxiv"
      ],
      "env": {
        "PUBMED_EMAIL": "[email protected]",
        "PUBMED_API_KEY": "your_key_here",
        "SEMANTIC_SCHOLAR_API_KEY": "your_key_here"
      }
    }
  }
}

You can run three MCP servers locally to query academic databases—arXiv, Semantic Scholar, and PubMed—and have an integrated assistant fetch, deduplicate, and summarize papers in real time. This MCP setup lets you issue natural language questions and receive structured results with abstracts, authors, and BibTeX citations from multiple sources, all coordinated through a single client and independent servers.

How to use

You use an MCP client to connect to three stdio MCP servers that run locally. Each server exposes a dedicated interface for searching and retrieving papers from a specific data source. You can perform multi-source searches, get detailed paper data, and generate BibTeX citations. The workflow is to issue a query, let the system search arXiv, Semantic Scholar, and PubMed, deduplicate results, and present a combined, ranked list with links, abstracts, and citations.

How to install

{
  "mcpServers": {
    "research-arxiv": {
      "command": "npx",
      "args": ["-y", "researchmcp", "arxiv"]
    },
    "research-semantic-scholar": {
      "command": "npx",
      "args": ["-y", "researchmcp", "semantic"]
    },
    "research-pubmed": {
      "command": "npx",
      "args": ["-y", "researchmcp", "pubmed"]
    }
  }
}

Additional sections

Prerequisites include Node.js 18 or newer. You also need an MCP-compatible client to connect to the servers.

Install steps are straightforward: install the three MCP servers via npx, then run them through your MCP client. If you prefer local development, you can run each server directly for testing.

Configuration is handled entirely through the MCP client configuration. The three servers shown above are designed to be used together to perform multi-source searches and to generate BibTeX citations for selected papers.

Security notes: No API keys are required to run these three servers out of the box. If you later add API keys for rate-limited sources, you can pass them via environment variables in your MCP client configuration.

Troubleshooting tips include verifying that each server starts correctly and that the MCP client can reach all three endpoints. Check that your Node.js version matches the prerequisites and that your environment allows the npx commands to execute.

Notes on configuration and usage

The three MCP servers support the following tools and actions: search and retrieve paper data from arXiv, Semantic Scholar, and PubMed, convert paper data to BibTeX, and provide detailed paper metadata such as titles, authors, abstracts, years, venues, and URLs.

Available tools

search_arxiv

Search arXiv by keyword, author, or subject with optional year filters and sorting to pull matching papers.

get_arxiv_paper

Retrieve detailed metadata for a specific arXiv paper by its ID.

arxiv_to_bibtex

Convert an arXiv paper entry into BibTeX format.

search_semantic_scholar

Search Semantic Scholar with a query and optional year filters to obtain papers and citation data.

get_semantic_scholar_paper

Fetch full details for a paper by Semantic Scholar ID or DOI.

get_paper_citations

Retrieve papers that cite a specific Semantic Scholar paper.

semantic_scholar_to_bibtex

Convert a Semantic Scholar paper into BibTeX format.

search_pubmed

Query PubMed for biomedical papers with optional year filters.

get_pubmed_paper

Get detailed information for a PubMed paper by PMID.

pubmed_to_bibtex

Convert a PubMed entry to BibTeX format.