Claude Code retrieves open PDFs for a 76-study review
Nam reports iterating with Claude Code on a working open-access PDF retrieval pipeline for a 76-study biopsy review, using API fallbacks while paywalled papers retained institutional access.
aperivue.com
Original prompt
Shared by the source author. See source notes for context.
I have a list of 76 DOIs. I need to download the full-text PDFs.
Reach people exploring AI agents
See placements →What happened
Nam reports iterating with Claude Code on a working open-access PDF retrieval pipeline for a 76-study biopsy review, using API fallbacks while paywalled papers retained institutional access.
Reported by Yoojin NamHow the agent did it
- Supply the included-study DOI list and limit retrieval to open-access papers.
- Test the initial script and report the PMC website access failure.
- Use Unpaywall and Europe PMC API retrieval with metadata fallbacks.
Where a person stepped in
Nam defines the review and iterates on real failures. Paywalled papers remain an institutional-access task, and the source does not independently audit retrieval completeness.
Yoojin Nam, a radiology resident, needs full-text PDFs for seventy-six studies in a review of CBCT-guided lung biopsies. She supplies the DOI list and asks Claude Code to handle the open-access portion first, separating it from papers that require institutional access.
The first Python implementation encounters a JavaScript challenge at PMC's website. Nam reports that three rounds of conversation produce a working retrieval chain: Unpaywall first, Europe PMC's REST API next, then OpenAlex and Crossref fallbacks. She saves the process as a reusable skill instead of rebuilding the same download method for the next research project.
The source describes a successful open-access pipeline, not unrestricted access to all seventy-six studies. Its time savings are the author's report; no independent completeness audit is supplied. The useful recovery is to communicate the actual access failure and use documented public APIs, while retaining institutional retrieval for unavailable full text. Nam explicitly supplies domain expertise and tests skills on actual manuscripts, correcting missed reporting items or inappropriate statistical choices. This entry stays with the PDF-retrieval case rather than treating the toolkit's other twenty-two skills as separate completed outcomes from the same article.
Prompt
Original prompt
Source: Blog · Yoojin Nam
Frequently asked questions
Were all seventy-six papers freely downloaded?
The source distinguishes open-access retrieval from the remaining paywalled papers requiring institutional access.
How was the PMC website failure handled?
After iteration, the pipeline used Europe PMC’s REST API alongside Unpaywall and metadata fallbacks.
Who supplies research expertise?
Nam supplies clinical and methodological knowledge and tests the skills against her actual research work.
More details
Remaining steps
- Save the working retrieval process as a reusable skill.
- Handle paywalled full text separately through institutional access.
About this account
Claude Code builds, tests and revises DOI-based retrieval tools, adapting to a website failure with public API fallbacks and preserving the successful method as a skill.
Firsthand account by Yoojin Nam; original publication date verified against Primary blog April 7, 2026. Outcomes are the author’s reports, not independently reproduced. Original prompt excerpt; the complete instruction explicitly limits the initial batch to open-access papers.
Credit: Yoojin Nam
Source published 7 Apr 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Python · DOI list · Unpaywall · Europe PMC REST API · OpenAlex · Crossref
Open-access retrieval · DOI batches · API failure recovery
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.