Grant and manuscript critique

Know what a reviewer will see.

A mock study section is the best check before you submit, and most investigators cannot get one on demand. Crucible reads your grant or manuscript against the actual review criteria and returns comments you can act on. It does not write your science.

Request access How it works

Crucible in about a minute

Earlier

Researchers who get substantive feedback early submit stronger applications — and Crucible makes that available at any stage.

On demand

Meaningful feedback whenever you need it — no waiting for a colleague's calendar or a mock study section to convene.

Iterative

Upload a revision when you are ready. Each round is a new read against the same criteria.

What Crucible does

The reviewer's perspective,
on demand.

Crucible reads what you have already written and shows you where a reviewer is likely to pause: a thin aim, a claim that outruns the data, a section that does not match the mechanism. Your science and your sentences stay yours.

Grant Applications

Scored on the NIH 1–9 scale against that mechanism's criteria (R01, K23, VA Merit, R41, and others). The number is a rough read of how many weaknesses showed up and how serious they are, not a prediction of your study section. Attend to what is flagged. Inline Word comments, filterable by severity, plus a downloadable summary.

Summary Statement Response

Upload the summary statement. Crucible builds weakness tables and a structured plan for the resubmission introduction and point-by-point responses. You write the response.

Manuscript Feedback

Qualitative critique of argument, evidence, framing, and gaps. No numeric score. Reporting-standard checks where they apply (CONSORT, STROBE, PRISMA, ARRIVE).

What you get

What a careful reviewer would actually give you.

Scored criteria, for grants

Each review factor gets a 1–9 score and a written strength/weakness note. The scores are approximate. They track the number and seriousness of the weaknesses, not a calibrated NIH result. Use the number to see where to look. The comments are what to act on. Manuscripts are not scored that way.

Annotated documents

A .docx back with comments on the paragraph they refer to. Filter Hi / Med / Lo.

Summary report

Scores, comments, and a timestamped footer.

Mechanism-specific review

R01, K23, VA Merit, R41, and others, each against its own criteria.

Another round

Revise, upload again, compare.

Whole-application check

Upload the assembled ASSIST application. Crucible reads Aims, Approach, Budget, Human Subjects, and Biosketches against each other and flags where they disagree. Optional AI text detection shows which sections to re-read.

The Crucible Philosophy

"Every word in the application remains yours. Crucible gives you the reviewer's perspective so you can change the work while it is still yours to change. It reads, critiques, scores, and comments. It does not write your science."

Crucible does not generate Specific Aims, or any other scientific text, for you to submit as your own. Fully aligned with NIH Policy NOT-OD-25-132.

How it works

Four steps, in the order the app actually uses.

Create a project

Create a project and choose the mechanism or manuscript type.

Upload

Upload the .docx or .pdf into that project.

Read the comments

Read the comments and the weaknesses they point to. Use the scores only as a map of where the problems clustered.

Revise

Revise the document and upload again.

Using Crucible on an R01, in a minute

Who it's for

Built for investigators who do not have a mock study section on tap.

Senior labs circulate drafts. Everyone else waits on a colleague's calendar, or submits without that read. Crucible is for that second group: K and other early-career awards, investigators at smaller programs, and anyone staring at a summary statement.

Early-career awards

K and other early-career awards. You need a careful read before you submit, and a mock study section is not sitting on your calendar.

Smaller programs

Investigators at smaller programs, where the colleague who would convene that section is not down the hall.

Resubmission

Anyone staring at a summary statement. Upload it and get a structured plan; you write the response.

Privacy

What we actually guarantee.

No training by default

Your grant or manuscript is not used to train models: not by us, and not by our model providers, unless you explicitly opt in.

NIH policy

Aligned with NIH NOT-OD-25-132: Crucible critiques. It does not generate Specific Aims, or any other scientific text, for you to submit as your own.

Unpublished work

Unpublished work is treated as unpublished work. We do not sell it or reuse it for other customers.