Check a manuscript before you submit

Paste or upload a draft — no DOI needed — and we report what it documents about reproducing its analysis: whether versions are pinned, whether the software environment is closed, whether the data and code are reachable. Free, private, and yours alone.

What this is not.

Nothing is executed. The manuscript is read by a language model and gives a ceiling: an upper bound on what a code check could confirm. The repository is not read here, so a lockfile or container named in the text stays a claim. Nothing here is applied, signed or published.

The manuscript

Paste the text

…or upload it PDF, Markdown or plain text. A PDF is read with poppler; line-break hyphens do not survive it, so a version like BWA-MEM 0.7.17 can come back mangled. Pasted text is the cleaner read.

DOI — optional If the manuscript already has one. Without it, the report is issued against the SHA-256 of the text you submit — a revised draft is a new subject, which is as it should be. Nothing is fetched from the DOI here; only the text you give us is read.

I agree that the manuscript text will be sent to Anthropic Claude Sonnet 5.5 running on Amazon Bedrock in the EU (Frankfurt, eu-central-1) to produce the reading of the manuscript. The result is returned to me alone, and the manuscript is discarded once the check finishes.

Model settings: effort high, which is Claude Sonnet 5.5's default and is not changed by this service, with adaptive thinking — the model decides how much to reason before it answers. The model, its effort and the assessor's version are recorded in the report.

Already published on bioRxiv or medRxiv? Check it from its DOI · How this rubric was applied to 201 bioRxiv papers