Happy slice of mcp pizzaMCP.pizza

pubmearch

MCP.Pizza Chef: Darkroaster

Aimed at medical researchers, this add-on runs a PubMed search using the site's own advanced search syntax, saves the results to a file on your computer, then counts the most common subject terms, tracks how they rise and fall over time, and reports how many papers appeared in each period. It only reads PubMed and never changes anything there. An NCBI account key is optional and merely lifts the request limit; a working email address is the one thing it refuses to start without.

Unmaintained · no code changes since May 2025; author calls it early-stage
Data
Web/Research

Use This MCP server To

See which topics dominate a research area this year Track how interest in a treatment shifted over five years Count how many papers appeared on a topic each quarter Save a literature search so I can revisit it later Get a quick summary of a field before writing a review

README

PubMed Analysis MCP Server

适用于PubMed的MCP server：这是一个刚刚开发的项目，功能仍在完善中，欢迎各位提出建议和改进！

Note: This is a newly developed project with features still being refined. Suggestions and improvements are welcome!

一个专业的PubMed医学文献分析MCP服务器，帮助科研人员快速洞察医学研究动态。

A professional MCP server for analyzing PubMed medical literature to help researchers quickly gain insights into medical research dynamics.

功能特点 / Features

  • 文献检索 / Literature Retrieval: 支持PubMed高级检索语法，可设置日期范围和结果数量。/ Supports PubMed advanced search syntax with date filtering.

  • 热点分析 / Hotspot Analysis: 统计关键词频率，识别热门研究方向，汇总相关文献。/ Analyzes keyword frequencies to identify popular research areas.

  • 趋势追踪 / Trend Tracking: 追踪关键词随时间的频率变化，揭示研究趋势演变。/ Tracks keyword changes over time to reveal evolving research trends.

  • 发文统计 / Publication Count: 提供灵活的时间周期设置，分析文献数量变化。/ Analyzes publication volume changes with customizable time periods.

  • 全面报告 / Comprehensive Reports: 一键生成包含热点、趋势和统计的分析报告。/ Generates complete reports with customizable parameters.

MCP工具 / MCP Tools

1. search_pubmed

搜索PubMed并保存结果。/ Search PubMed and save results.

主要参数 / Key parameters: 一般而言不需要显性设置，与大模型沟通即可。/ Generally, no need to set explicitly, communicate with large models.

  • advanced_search: PubMed搜索查询（必填，与高级检索语法相同）/ PubMed search query (required, same as advanced search syntax)
  • start_date: 开始日期（格式：YYYY/MM/DD）/ Start date (format: YYYY/MM/DD)
  • end_date: 结束日期（格式：YYYY/MM/DD）/ End date (format: YYYY/MM/DD)
  • max_results: 最大结果数（默认：1000）/ Maximum results (default: 1000)

2. list_result_files

列出可用的结果文件。/ List available result files.

3. analyze_research_keywords

pubmearch FAQ

Do I really need a key from NCBI?
No. The key is optional — the code simply warns and carries on without it. What it does insist on is an email address for NCBI; without that, searching refuses to run.
Is this still maintained?
The code has not changed since May 2025, and the author describes it as an early project that may be slow to receive updates.
Can I use this to find the hot topics in my field right now?
Yes — give it a PubMed search and a date range, and it returns the most frequent subject terms along with how their frequency has shifted.
Does it change anything in PubMed?
No, it only reads. It does write result files onto your own computer, inside the folder where the tool is installed.
Why does nothing happen when I ask for an analysis?
The two steps are separate. Run a search first, which saves a file, then point the analysis at that saved file by name.
There is a shortcut command called pubmearch_mcp — should I use it?
No. That shortcut points at a start-up function that does not exist in the code, so it fails. Use the module form shown in the README instead.
How many papers will it fetch?
Up to a thousand by default, and that limit is genuinely honoured. You can ask for fewer if you want a faster answer.
How hard is the setup?
Developer-level. You download the project, install it with uv or pip, then paste a settings block that points at the folder on your disk.
Which apps does it work in?
The README shows a Windows setup for Trae and mentions Cursor. Other apps that launch local tools use the same kind of settings block, but they are not documented here.