Paper Search MCP Server

Instructions

USE THIS TOOL WHEN:
- Searching for medical, clinical, or biomedical research
- You need peer-reviewed published papers (not preprints)
- Searching for drug studies, clinical trials, disease research

DOMAIN: Medicine, Biology, Pharmacology, Public Health, 
Clinical Research, Genetics, Biochemistry.

LIMITATION: PubMed provides metadata/abstracts ONLY, not full PDFs.

WORKFLOW FOR FULL TEXT:
1. search_pubmed(query) -> get DOI from results
2. download_scihub(doi) -> download PDF (if published before 2023)

Args:
    query: Medical/scientific terms (e.g., 'cancer immunotherapy', 'COVID-19 vaccine').
    max_results: Number of results (default: 10).

Returns:
    List of paper dicts with: paper_id (PMID), title, authors, 
    abstract, published_date, doi, url.

Example:
    search_pubmed("CRISPR gene therapy", max_results=5)

Implementation Reference

• paper_find_mcp/server.py:231-260 (handler)

  MCP tool handler for 'search_pubmed'. Decorated with @mcp.tool() for registration and schema definition via signature/docstring. Delegates execution to the generic _search function with 'pubmed' searcher.

  @mcp.tool()
  async def search_pubmed(query: str, max_results: int = 10) -> List[Dict]:
      """Search biomedical literature on PubMed (NCBI database).
      
      USE THIS TOOL WHEN:
      - Searching for medical, clinical, or biomedical research
      - You need peer-reviewed published papers (not preprints)
      - Searching for drug studies, clinical trials, disease research
      
      DOMAIN: Medicine, Biology, Pharmacology, Public Health, 
      Clinical Research, Genetics, Biochemistry.
      
      LIMITATION: PubMed provides metadata/abstracts ONLY, not full PDFs.
      
      WORKFLOW FOR FULL TEXT:
      1. search_pubmed(query) -> get DOI from results
      2. download_scihub(doi) -> download PDF (if published before 2023)
      
      Args:
          query: Medical/scientific terms (e.g., 'cancer immunotherapy', 'COVID-19 vaccine').
          max_results: Number of results (default: 10).
      
      Returns:
          List of paper dicts with: paper_id (PMID), title, authors, 
          abstract, published_date, doi, url.
      
      Example:
          search_pubmed("CRISPR gene therapy", max_results=5)
      """
      return await _search('pubmed', query, max_results)

• paper_find_mcp/academic_platforms/pubmed.py:155-215 (helper)

  Core implementation of PubMed search logic in PubMedSearcher.search(). Uses NCBI E-utilities API: esearch.fcgi to get PMIDs, efetch.fcgi to fetch details, parses XML to extract title, authors, abstract, DOI, etc., into Paper objects.

  def search(self, query: str, max_results: int = 10) -> List[Paper]:
      """搜索 PubMed 论文
      
      Args:
          query: 搜索关键词，支持 PubMed 查询语法
                 例如: "cancer[Title]", "Smith J[Author]"
          max_results: 最大返回数量（上限 10000）
          
      Returns:
          List[Paper]: 论文列表
      """
      # Step 1: 搜索获取 PMIDs
      search_params = {
          **self._get_base_params(),
          'term': query,
          'retmax': min(max_results, 10000),  # NCBI 限制
          'retmode': 'xml'
      }
      
      search_response = self._make_request(self.SEARCH_URL, search_params)
      if not search_response:
          return []
      
      try:
          search_root = ET.fromstring(search_response.content)
          ids = [id_elem.text for id_elem in search_root.findall('.//Id')]
      except ET.ParseError as e:
          logger.error(f"Failed to parse search response: {e}")
          return []
      
      if not ids:
          logger.info(f"No results found for query: {query}")
          return []
      
      # Step 2: 获取论文详情
      fetch_params = {
          **self._get_base_params(),
          'id': ','.join(ids),
          'retmode': 'xml'
      }
      
      fetch_response = self._make_request(self.FETCH_URL, fetch_params)
      if not fetch_response:
          return []
      
      try:
          fetch_root = ET.fromstring(fetch_response.content)
      except ET.ParseError as e:
          logger.error(f"Failed to parse fetch response: {e}")
          return []
      
      # Step 3: 解析论文数据
      papers = []
      for article in fetch_root.findall('.//PubmedArticle'):
          paper = self._parse_article(article)
          if paper:
              papers.append(paper)
      
      logger.info(f"Found {len(papers)} papers for query: {query}")
      return papers

• paper_find_mcp/server.py:95-113 (helper)

  Generic helper function _search that retrieves the platform-specific searcher (PubMedSearcher for 'pubmed'), calls its search method, and converts Paper objects to dicts for the tool response.

  async def _search(
      searcher_name: str, 
      query: str, 
      max_results: int = 10,
      **kwargs
  ) -> List[Dict]:
      """通用搜索函数"""
      searcher = SEARCHERS.get(searcher_name)
      if not searcher:
          logger.error(f"Unknown searcher: {searcher_name}")
          return []
      
      try:
          papers = searcher.search(query, max_results=max_results, **kwargs)
          return [paper.to_dict() for paper in papers]
      except Exception as e:
          logger.error(f"Search failed for {searcher_name}: {e}")
          return []

• paper_find_mcp/server.py:75-85 (registration)

  Global SEARCHERS dictionary instantiates singleton PubMedSearcher instance used by all pubmed tools.

  SEARCHERS = {
      'arxiv': ArxivSearcher(),
      'pubmed': PubMedSearcher(),
      'biorxiv': BioRxivSearcher(),
      'medrxiv': MedRxivSearcher(),
      'google_scholar': GoogleScholarSearcher(),
      'iacr': IACRSearcher(),
      'semantic': SemanticSearcher(),
      'crossref': CrossRefSearcher(),
      'repec': RePECSearcher(),
  }