PubMed & Europe PMC to Markdown (abstracts + OA full text)

Turn PubMed and Europe PMC search results into LLM-ready Markdown. Give it a search query or a list of PMIDs, PMCIDs or DOIs. Each article comes back as one Markdown file with YAML front matter: title, authors, journal, date, abstract, MeSH terms, keywords, citation count and license. For open-access papers you also get the full text: sections, tables as GitHub-flavoured Markdown, formulas, figure captions and the numbered reference list.

• Price: $0.002 per article record (metadata + abstract), $0.004 per open-access full-text article, plus $0.005 per run. 1,000 abstracts cost about $2.
• Official API only. All data comes from the Europe PMC REST API, which covers all of PubMed/MEDLINE plus PubMed Central, preprints and more. No HTML scraping, no proxies, no blocking.
• Built for RAG, AI agents and literature reviews. Drop the files into a vector store, an Obsidian vault or a Claude Code / LLM wiki, or read the dataset from the Apify MCP server.

Output example (one article, shortened)

---
title: 'Evaluating large language model clinical reasoning in glaucoma using retrieval-augmented generation'
authors: ['Yin H', 'Miao Q', 'Zhou W', 'Hu C', 'Guo Y', 'Shen L', 'Cai H', 'Grzybowski A', 'Jin K']
journal: 'Advances in ophthalmology practice and research'
date: 2026-05-26
pmid: '42666552'
pmcid: 'PMC13522369'
doi: '10.1016/j.aopr.2026.05.003'
keywords: ['glaucoma', 'Clinical Decision Support', 'Large Language Models', 'Retrieval-augmented Generation']
license: 'cc by-nc-nd'
open_access: true
full_text: true
---

# Evaluating large language model clinical reasoning in glaucoma using retrieval-augmented generation

## Abstract

**Background.** Large language models (LLMs) demonstrate strong performance in knowledge-based medical tasks, yet ...

## 1 Introduction

...

**Table 1.** ...

| ... | ... |
| --- | --- |
## References

1. ...

What does PubMed & Europe PMC to Markdown do?

It searches the biomedical literature the same way you would on PubMed or europepmc.org, then exports each hit as clean, structured Markdown plus JSON fields. It is useful when you need the text of the papers themselves, not only a list of titles:

• Abstracts for every PubMed record, with structured-abstract labels kept (Background, Methods, Results, Conclusions).
• Full text for open-access articles (several million papers in PubMed Central), converted from the publisher's JATS XML: section headings with numbering, paragraphs, bullet lists, tables as Markdown tables, display formulas as $$ ... $$ blocks, figure and table captions, acknowledgments and the reference list.
• Metadata for filtering and citation: PMID, PMCID, DOI, authors, ORCID iDs, affiliation, journal and ISO abbreviation, volume, issue, pages, publication types, MeSH terms (and major topics), chemicals, grants, license, Europe PMC citation count, and links to PubMed, PMC, Europe PMC and doi.org.
• Scheduled monitoring: with a state key, a weekly run returns only papers it has not exported before, so you can keep a topic feed or a living review up to date.

Why use it?

• RAG over medical literature: every item has a single markdown_file field ready to chunk and embed, with front matter you can use as metadata filters (year, MeSH, license, open access).
• AI agents and MCP clients: an agent can call the Actor with a query and read back full papers as text, instead of fighting PDF parsing.
• Systematic and scoping reviews: export all hits of a search strategy with abstracts and MeSH for screening, then pull full texts of the included studies by PMID.
• Knowledge bases: build an Obsidian or Notion library of papers on a topic, one Markdown file per paper.
• Pharma, biotech and health-tech research: track new publications on a target, drug or device every week.

How to use it

1. Click Try for free and open the Input tab.
2. Type a search query, for example "large language model" AND clinical, or paste PMIDs, PMCIDs or DOIs into the identifier list.
3. Set Max articles (this caps your cost) and, if you want only papers with full text, tick Open-access articles only.
4. Click Start. A run of 10 articles usually finishes in under 10 seconds.
5. Open the Output tab: the Articles view is a table, the Markdown files view shows the files. Download as JSON, CSV, Excel or HTML, or fetch the dataset through the Apify API.

Input

Field	Type	Default	Notes
query	string		Europe PMC search syntax: keywords, "phrases", AND / OR / NOT, fields such as TITLE:, AUTH:, JOURNAL:, MESH:, PUB_TYPE:"review".
ids	string[]	[]	PMIDs, PMCIDs, DOIs, or PubMed / PMC / Europe PMC / doi.org URLs. Can be combined with a query.
maxResults	integer	100	Maximum articles stored. Each stored article is one billing event.
fullText	boolean	true	Convert open-access full text.
openAccessOnly	boolean	false	Only return open-access articles.
fromDate / toDate	string		YYYY or YYYY-MM-DD (first publication date).
sort	string	relevance	relevance or newest.
includeReferences	boolean	true	Append the reference list to the full text and return it in references.
stateKey	string		Named store that remembers exported articles, for scheduled runs.
maxConcurrency	integer	3	Parallel requests to Europe PMC (max 8).

Example inputs:

{ "query": "\"retrieval augmented generation\" AND clinical", "maxResults": 50, "openAccessOnly": true }

{
    "query": "MESH:\"Diabetes Mellitus, Type 2\" AND PUB_TYPE:\"review\"",
    "fromDate": "2025",
    "sort": "newest",
    "maxResults": 200,
    "fullText": false
}

{ "ids": ["31452104", "PMC13340859", "10.2196/86073"] }

Output

One dataset item per article. Main fields:

Field	Description
title, authors, journal, pub_date, pub_year	Bibliographic data
pmid, pmcid, doi, source	Identifiers (source is MED for PubMed, PPR for preprints, ...)
abstract_markdown	Abstract as Markdown
mesh_terms, mesh_major_topics, keywords, chemicals, grants, pub_types	Indexing
is_open_access, license, cited_by_count	Access and impact
full_text, full_text_markdown, section_titles, table_count, figure_count	Converted full text (open access only)
references, reference_count	Reference list (label, text, doi, pmid)
markdown_file	Front matter + title + abstract + full text, ready to save as .md
word_count, content_hash, extraction_error, fetched_at	Housekeeping

A run summary (matches, stored, full texts, API requests, charged events) is saved as SUMMARY.json in the run's key-value store.

How much does it cost to export PubMed articles?

Pay per event, no subscription:

Event	Price
Run start	$0.005
Article record (metadata + abstract)	$0.002
Open-access full-text article	$0.004

Examples: 100 abstracts = $0.205. 500 open-access full texts = $2.005. Each article is charged once, never both events. Open-access bodies shorter than 500 words (conference abstracts, abstract-only deposits) are still returned but charged as a record. Platform usage is small because the Actor only calls a JSON/XML API. Set Max articles or a maximum cost per run to cap spending; the Actor stops cleanly when the budget is reached.

Tips

• Want only papers you can read in full? Tick Open-access articles only, or add OPEN_ACCESS:y to the query.
• Screening first, full text later: run with fullText: false for cheap abstracts, then pass the included PMIDs with fullText: true.
• Precise searches: Europe PMC supports field search, e.g. TITLE:"sepsis" AND AUTH:"Seymour CW", JOURNAL:"Lancet", FIRST_PDATE:[2025-01-01 TO 2025-06-30].
• Weekly feed: schedule the Actor with sort: newest, a fromDate, and a stateKey.

FAQ and limitations

• Is it legal? The Actor uses the public Europe PMC REST API, the same service EMBL-EBI provides for text mining. Abstracts and metadata are made available for reuse by NLM and Europe PMC; full text is only fetched for articles in the open-access subset, and each item carries its license so you can respect reuse terms (for example CC BY-NC).
• Why no full text for some articles? Only open-access articles deposited in PubMed Central have full text in the API. Others return metadata and abstract, with the reason in extraction_error.
• Figures are exported as captions plus the figure file name, not as images.
• Formulas use the article's TeX when available; otherwise MathML is converted to readable TeX-like text, which can differ slightly from the typeset version.
• Questions, bugs or a field you need? Open an issue on the Issues tab and we will reply.