Docs
REST API for retrieving machine-generated screening signals, per-axis availability, and recorded run provenance.
Base URL
https://preprints.aiAll endpoints below are relative to this base.
Endpoints
GET /v1/stats
Aggregate counts for the live pipeline: indexed papers, generated signals, queue state, available-axis distributions, and complete-run provenance.
curl https://preprints.ai/v1/statsDistributions exclude axes whose required provenance is unavailable. Use the accompanying coverage counts as the denominator.
GET /v1/pipeline
Recent screening signals, canonical screening_signal objects, and filter heatmaps built only from displayable axes.
curl https://preprints.ai/v1/pipelineGET /v1/papers
Paginated catalogue. Compose evidence, trust, novelty_min, novelty_max, q, and complete_only. Axis filters exclude rows where that axis is unavailable.
curl "https://preprints.ai/v1/papers?trust=A&novelty_max=4&complete_only=true"GET /v1/score/{identifier}
Full screening report for an assessed DOI, arXiv ID, or internal paper ID.
curl https://preprints.ai/v1/score/10.1101/2025.01.15.633214POST /v1/assess
Request a screening run for a specific paper. New external submissions are currently limited to private-preview clients; public catalogue and report reads remain open.
curl -X POST https://preprints.ai/v1/assess \
-H "content-type: application/json" \
-d '{"doi": "10.1101/2025.01.15.633214"}'Screening contract
A compact profile such as AC4 means Evidence A, Trust C, Novelty 4. It is a set of machine-generated screening signals—not peer review, scientific verification, or a publication verdict. Consumers should read screening_signal.axes before displaying any legacy grade field.
| Axis | Scale | Question |
|---|---|---|
| Evidence | A–E | Rubric-guided model screening of methods and claim support. |
| Trust | A–E | Recorded transparency, reproducibility, and reporting markers. |
| Novelty | 1–10 | Machine-generated distinctness signal; lower is more novel. Read axes.novelty.basis to distinguish literature-grounded, reconstructed, and LLM-derived findings. |
Evidence bands
| Letter | Meaning | Integrity score range |
|---|---|---|
| A | Few material concerns detected under the current rubric | ≥ 0.85 |
| B | Generally strong signal with minor detected limitations | 0.70–0.84 |
| C | Mixed signal with notable limitations to inspect | 0.55–0.69 |
| D | Substantial concerns detected | 0.40–0.54 |
| E | Critical concerns detected under the current rubric | < 0.40 |
Run and axis status
screening_signal.status is complete, limited, or withheld. Each axis is separately available, limited, or unavailable, with machine-readable limitation codes. Novelty also includes a basis and human-readable label; a reconstructed or LLM-derived value remains visible as a limited machine-generated signal. provenance_complete: true confirms execution provenance and score contracts only; scientific_verification is always false.
Rate limits
Public reads require no authentication. The default anonymous throttle is 30 requests per minute per client; authenticated partner limits may differ. Ask for a data export rather than scraping at high volume.
Webhooks
Not yet public. If you're building an integration that needs push-style delivery of new screening signals, contact us — we have a private beta.
Machine-readable specs
Full OpenAPI 3.1 spec at /openapi.json. LLM-friendly summary at /llms.txt.
Methodology
For the current pipeline architecture, score axes, hard gates, and limitations, see the methodology page.
Stability
The /v1 API is stable. New fields will be added additively; existing fields will not change shape within v1. Breaking changes go to /v2.
Source & transparency
Assessment provenance records the manuscript input, model and panel metadata, automated-check results, and confidence signals where available. Model sampling means a rerun is not guaranteed to produce the same result; stability and calibration must be measured rather than assumed.