Find the references your paper is missing.

Upload a draft — the pipeline flags statements that need support and recommends real papers from live academic databases. Every suggestion is retrieval-grounded: the model only selects and ranks, it never invents a reference.

S0 parse & strip refsS1 select statements S2 search up to 9 sourcesS3 listwise rerank S3.5 snowballing (optional)S4 dedup & match S5 report

Upload a paper

Drop your paper here, or browse
.md · .txt · .pdf · up to 25 MB — a References section in your draft is detected automatically and used to separate “already cited” from “newly suggested.”
Advanced options
Explore how it works

Analyzing your paper

S0
Parse
S1
Select
S2·3
Search & rank
S4
Dedup
S5
Report

Results

How it works

Built so a fake citation can't exist

You may have seen chatbots invent plausible-looking papers. This tool closes that door by design: it searches nine scholarly databases — Semantic Scholar, OpenAlex, Crossref, arXiv and more — and the AI's only job is to choose and rank among what the search actually returned. Under the hood, that discipline is engineered: a deterministic pipeline, schema-validated model calls, per-source rate governance, resumable on-disk artifacts.

Every paper is real — and checkable

Recommendations come from live database searches, never from the AI's memory. Each one carries its DOI or arXiv link, so you're one click from the actual paper — and every score comes with a stated reason you can weigh for yourself.

It knows what not to flag

Common knowledge, your own findings and pure derivations are explicitly off-limits, and each section has a flagging budget — citation-dense ones like Related Work get more headroom. Your draft comes back annotated, not carpeted.

Infrastructure, not a wrapper

For the technically curious: a deterministic S0→S5 workflow whose model calls are schema-validated with automatic retry (via agentmaker), nine search adapters behind one protocol, per-source throttles honouring each API's etiquette, and stage artifacts that make runs resumable, diffable and fully offline-testable.

The pipeline from draft to report, stage by stage

S0
Read & set aside Your draft is split into sections; PDFs are parsed page by page. If a reference list is present it's set aside — it must not bias what gets flagged, and it becomes the answer key for "already cited" later.
S1
Flag sentences that need support AI The AI reads each section and flags the sentences that lean on prior work — each with a reason, a confidence score and ready-made search terms. Every flag quotes your text verbatim, so it's traceable to an exact spot in the draft.
S2
Search nine databases at once Each flagged sentence is searched across up to nine indexes, phrased in each engine's idiom — keyword engines get distilled keywords, semantic engines get the whole sentence, hybrid engines get both. Duplicate hits merge by DOI and arXiv id, then by fuzzy title-and-author match.
S3
Shortlist, with reasons AI All of a sentence's candidates are judged together — one call, one comparable 0–10 scale. The top few survive, each with a one-line why; a paper published after yours can't be your source, so its score is capped.
S3.5
Follow the references AI optional Keyword search favours papers that speak today's vocabulary; the classics often don't. This step reads the top hits' own reference lists, finds the paper they all cite, and lets it compete in a second ranking — strictly additive, so it can add results but never remove one.
S4
Merge & check against your bibliography A paper supporting three sentences becomes one entry, ranked for breadth as well as strength (best score × ln(1 + sentences supported)). Anything already in your reference list is tagged already cited instead of being pitched back to you.
S5
Report A readable report plus a structured JSON — new suggestions first, already-cited confirmations listed separately, and every intermediate result kept for inspection.

Design decisions fair questions, honest answers

01 Could it invent a paper, like chatbots do?

No — not won't, can't. The AI never writes a citation; it returns index numbers into the list the search produced. A number outside the list is discarded by code, and an unscored candidate defaults to zero. Fabrication isn't discouraged — it's unrepresentable.

02 Why does a run take a few minutes?

Courtesy, mostly. arXiv permits one request every three seconds; other databases meter by the second, and each waits in its own queue so it never delays the rest. Brute-force parallelism can't help — quotas are enforced server-side, so ten connections would just collect ten refusals. What I have done: merging a sentence's queries into one OR-request, which cuts the wait several-fold without losing a single result.

03 Can a 0–10 score from an AI be trusted?

Only if scores share a scale — so all of a sentence's candidates are judged in a single look rather than one at a time, which would cost ~30× the calls and produce scores that drift between them. Every score also states its reason, so you can hold it against the abstract yourself.

04 Will it find the 20-year-old classic my field cites?

Keyword search often won't — older papers speak yesterday's vocabulary. The snowballing step reads your best matches' reference lists and surfaces the paper they cite in common, matching identities across DOI and arXiv records so one classic isn't split into two half-counts. Live case: the original RAG paper, missed by three rounds of search, recovered and re-scored 10/10.

05 What if a database is down that day?

The run bends; it doesn't break. A failing source costs only its own results, a corrupted cache entry counts as a miss, and an error on one sentence can't discard the finished work of the others. You still get the full report from whatever answered.

06 How do you know it actually works?

Because it's built to be measured. A cutoff date — enforced at the API and re-checked locally — lets you replay history: strip a published paper's bibliography and count how much gets recovered. Snowballing ships off by default to keep that baseline clean, and every stage writes its output to disk, so any two runs can be diffed line by line.

Nine sources you already know each kept to its official request etiquette

OpenAlex 50 ms Crossref 120 ms Europe PMC 250 ms Semantic Scholar 1.1 s DBLP 1.5 s arXiv 3 s Exa 120 ms key CORE 2.5 s key Lens 6 s key

Six need no key at all; Exa, CORE and Lens join automatically when a key is configured. The number on each pill is that source's enforced minimum gap between requests — a slow source queues itself, never the others, and a failing one costs only its own results.

The pipeline runs right in front of you — five stages, live progress, and a report you can take with you.

Try it with your draft