Run a literature review workflow with Claude Code

A physician-researcher used Claude Code to turn a research question into a screened, cited, and drafted review.

Done with Claude Code· Advanced

What they did
Dalton Haslam, MD MBA, created Claude Code skills with a Python wrapper around free literature APIs. The agent first helps shape the question into PICO form, then pulls results from PubMed, OpenAlex, and Semantic Scholar, removes repeats, checks abstracts twice, traces references, downloads open-access PDFs, assesses quality, and produces the write-up. Six approval points are built in, and the user handles any paywalled PDFs manually.
How it went
The repository contains a sample review on congestive heart failure care transitions, plus the full output bundle. No speed or accuracy metrics are provided, and the project warns users to confirm the results themselves.
Tips
Expect the process to stretch across about 4–6 sessions. A free Semantic Scholar key helps avoid shared-pool rate limits, and the manual PDF step is designed for macOS.

Try it yourself with Claude Code

I want to produce a systematic review for this research question: [research question]. Start by asking me the minimum set of questions needed to frame it as PICO. Then search [sources such as PubMed, OpenAlex, and Semantic Scholar], remove duplicates, review titles and abstracts using these inclusion rules: [criteria], and trace citations from the most relevant papers. Save any open-access PDFs you can retrieve into [folder]. Before any irreversible action, including sending requests, submitting text, deleting files, or paying for access, stop and ask for my approval. Also pause at each of the six checkpoints and wait for me to continue. When the review is complete, draft a synthesis that lists the included studies, summarizes the findings, and records what was left out and why.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.