# CitePulse

> Academic citation analytics for AI research agents, grant offices, and PIs — the free-rebuild layer under Web of Science ($212k/yr) and Scopus (~$140k/yr). Bibliography verification with retraction detection, paper/author/institution/journal metrics, rising-topic and funder-impact scans, and grounded literature briefs — all via x402 micropayments on Base. Built entirely on free, open scholarly infrastructure: OpenAlex, Crossref, Semantic Scholar, Unpaywall.

## What This Is

CitePulse is a pay-per-query API that verifies citations and surfaces research-impact intelligence deterministically from open scholarly databases, with Claude synthesis layered only on top of already-verified facts (never used to decide whether a citation exists or is retracted).

## Payment

All endpoints require x402 micropayment (USDC on Base mainnet).
Protocol: x402 | Network: eip155:8453 | Asset: USDC
Payment address: 0x50ab2018c06c6E4eAA9BA52057Eb55eD284912fc

To use: send a request without a PAYMENT-SIGNATURE header to receive a 402 response with payment requirements. Sign the payment and retry with the PAYMENT-SIGNATURE header.

## Endpoints

GET/POST /api/ref-check?citations={doi1,doi2,...} (flagship)
- Verify up to 20 citations (DOIs or free-text references) in one call
- Returns per-reference verdict: verified | retracted | title_year_author_mismatch | not_found | low_confidence
- Retraction/withdrawal detection cross-checks Crossref's Retraction Watch feed (the `updated-by` relation) AND OpenAlex's `is_retracted` flag
- Mismatch detection compares claimed title/authors/year (if supplied) against the resolved real record
- POST body: {"citations": [...]} — items can be plain strings or {doi, title, authors, year} objects for stricter checking
- Deterministic — Crossref + OpenAlex decide existence/retraction, never an LLM guess
- Price: $0.15

GET /api/paper?doi={doi}|title={title}
- Single-paper lookup — citation count, influence, open-access status + best legal OA link (Unpaywall), venue, TLDR (Semantic Scholar)
- Deterministic passthrough, no synthesis
- Price: $0.05

GET /api/author?name={name}|openalex_id={id}
- Works count, total citations, h-index, i10-index, 10-year output/citation trend, affiliation timeline
- disambiguation_confidence flag (high/medium/low) — surfaces honestly when a name search spans multiple plausible researchers (a known OpenAlex weakness, especially for transliterated Chinese/Korean names)
- Price: $0.10

GET /api/institution?name={name}&compare_to={name}
- Institution research-output benchmark — works, citations, h-index/i10-index, 10-year trend, topic-strength breakdown
- Optional compare_to for a two-way side-by-side benchmark
- Price: $0.15

GET /api/journal?name={name}
- Journal/venue metrics (works, citations, DOAJ status, APC) + Tavily-grounded, cited legitimacy signals
- Never renders a bare "predatory"/"legitimate" verdict — evidence list + neutral framing only
- Price: $0.10

GET /api/topic-scan?topic={topic}&years={n}
- Rising-topic scan for a research field — what's rising, what's foundational, grounded in top-cited OpenAlex works
- Price: $0.20

GET /api/funder-impact?funder={name}&years={n}
- What a funder's grants have produced — highest-impact funded works, field distribution, grounded in OpenAlex's grants.funder linkage (a lower bound, not every funded paper carries structured grant metadata)
- Price: $0.20

GET /api/literature-brief?question={question}
- Grounded state-of-research synthesis for a specific question, citing only works retrieved from OpenAlex/Semantic Scholar
- Not a substitute for a systematic review — explicitly framed as a starting point
- Price: $0.25

## Discovery

- OpenAPI spec: https://citepulse.theaslangroupllc.com/openapi.json
- Agent discovery: https://citepulse.theaslangroupllc.com/.well-known/agent.json

## Notes for AI Agents

- All endpoints return structured JSON with a `disclaimer` field
- ref-check, paper, author, and institution are fully deterministic — no LLM in the existence/metrics path
- journal, topic-scan, funder-impact, and literature-brief use Claude to synthesize a narrative READ over already-verified facts; every cited work must come from the retrieved data, never invented
- Semantic Scholar enrichment (citation counts, TLDRs) is best-effort — its keyless tier rate-limits under load; a `coverage.semantic_scholar` field reports "ok" or "degraded" so you know when it was omitted
- OpenAlex direct ID/DOI lookups are free of rate-limit cost; only free-text search fallbacks draw against a shared daily credit ceiling
- Non-English literature, preprints, and grey literature have measurably weaker coverage in all underlying sources — every endpoint states this as a coverage caveat rather than hiding it
- Author/institution identity resolution is best-effort disambiguation, not a verified legal identity match