OpenCitations Scraper

Retrieve open scholarly citation links, outgoing references, citation counts, and bibliographic metadata from OpenCitations by DOI, PMID, or OpenCitations Meta ID (OMID).

The Actor uses the official OpenCitations Index v2 and Meta v1 REST services. It is useful for literature reviews, citation-network analysis, research intelligence, bibliometrics, academic knowledge graphs, and RAG pipelines that need structured citation data without scraping publisher pages.

What it extracts

The dataset always has one stable shape. Fields that do not apply to the selected operation are returned as null or an empty array, which makes exports and agent workflows predictable.

Field	Description
recordType	citation, metadata, or counts discriminator
queryIdentifier	Normalized DOI, PMID, or OMID used for the request
oci	Open Citation Identifier for a citation relationship
citingIdentifiers / citedIdentifiers	All persistent IDs OpenCitations attaches to each work
citingDoi / citedDoi	Convenient DOI values when available
creationDate / timespan	Citing-work date and XSD duration between publications
journalSelfCitation / authorSelfCitation	Self-citation flags when known
title, authors, publicationDate	Core bibliographic metadata
venue, volume, issue, pages	Publication context
workType, publisher, editors	Additional bibliographic details
citationCount / referenceCount	Incoming and outgoing totals in count mode
sourceUrl, scrapedAt	Exact REST endpoint and retrieval timestamp

When to use it

Use this Actor when you already know one or more persistent scholarly identifiers and want citation relationships or bibliographic data. Common workflows include:

• build an incoming-citation list for a paper;
• extract a paper's bibliography as outgoing citation edges;
• compare citation and reference counts across a DOI portfolio;
• resolve OpenCitations metadata for known works;
• generate nodes and edges for a scholarly knowledge graph;
• feed structured, source-linked records to an AI research agent.

This Actor is not a full-text academic search engine and does not search papers by arbitrary keywords, scrape Google Scholar, download articles, or bypass publisher access controls. OpenCitations covers open citation data contributed by its source indexes, so results may differ from proprietary citation databases.

Input

Input	Type	Default	Description
operation	string	citations	citations, references, both, metadata, or counts
identifiers	string[]	example DOI	Up to 50 DOIs, PMIDs, or bibliographic OMIDs
maxResults	integer	1000	Global cap after normalization and deduplication; Free-plan users receive at most 5 records
requestTimeoutSeconds	integer	30	Per-request timeout from 5 to 60 seconds

DOIs can be supplied as bare values, with a doi: prefix, or as https://doi.org/... URLs. A numeric identifier is interpreted as a PMID. Explicit examples include:

{
  "operation": "both",
  "identifiers": [
    "10.7717/peerj-cs.421",
    "pmid:33817056",
    "omid:br/06101801781"
  ],
  "maxResults": 250,
  "requestTimeoutSeconds": 30
}

Operation behavior

• citations returns incoming citation edges: works that cite each requested work.
• references returns outgoing citation edges: works cited by each requested work.
• both runs both directions and labels every row with its concrete operation.
• metadata returns one bibliographic metadata record when OpenCitations Meta resolves the identifier.
• counts returns one record containing both incoming citation and outgoing reference totals.

The maxResults cap applies to the entire run, not separately to every input. Apify Free-plan users are limited to 5 delivered records per run; paid-plan users can use the requested cap up to 10,000. When the free-tier cap is reached, the Actor exits gracefully with a clear status message and records the applied limit in SUMMARY. Duplicate identifiers are normalized and removed before requests start. Citation records are also deduplicated by operation, query identifier, and OCI.

Output example

Citation mode returns records like this:

{
  "recordType": "citation",
  "operation": "citations",
  "queryIdentifier": "doi:10.7717/peerj-cs.421",
  "oci": "06404435594-06101801781",
  "citingIdentifiers": [
    "omid:br/06404435594",
    "doi:10.1063/5.0110347",
    "openalex:W4322579315"
  ],
  "citedIdentifiers": [
    "omid:br/06101801781",
    "doi:10.7717/peerj-cs.421",
    "openalex:W3134956838",
    "pmid:33817056"
  ],
  "citingDoi": "10.1063/5.0110347",
  "citedDoi": "10.7717/peerj-cs.421",
  "creationDate": "2023",
  "timespan": "P1Y",
  "journalSelfCitation": false,
  "authorSelfCitation": false,
  "workIdentifiers": [],
  "title": null,
  "authors": [],
  "publicationDate": null,
  "venue": null,
  "volume": null,
  "issue": null,
  "pages": null,
  "workType": null,
  "publisher": null,
  "editors": [],
  "citationCount": null,
  "referenceCount": null,
  "sourceUrl": "https://api.opencitations.net/index/v2/citations/doi:10.7717/peerj-cs.421",
  "scrapedAt": "2026-08-28T12:00:00.000Z"
}

Every run also writes a SUMMARY key-value record with the outcome class (DATA, EMPTY, BLOCKED, FAILED, or REJECTED), successful and failed request counts, delivered record count, cap state, and warnings. Diagnostics are kept out of the business-entity dataset.

Pricing

The Actor uses transparent pay-per-event pricing. Apify charges synthetic events automatically; the code does not manually charge them.

Event	Price	When charged
Actor start	$0.00005	Once when a run starts
OpenCitations result	$0.00060	Once per schema-valid default-dataset record

For example, a paid-plan run returning 100 records costs $0.06005 in event charges. A valid no-results run costs only the $0.00005 start event. A count request normally returns one record, for $0.00065 total. A Free-plan run returns at most 5 records, for a maximum event charge of $0.00305. Platform usage is included in these event prices rather than charged separately; your configured maximum total charge still applies.

Reliability, limits, and data provenance

OpenCitations documents a fair-use rate limit of 180 API requests per minute per IP. This Actor stays below that threshold with a minimum interval between requests, retries transient connection errors and server failures, honors Retry-After on rate limits, and stops after three attempts. Count mode uses two API requests per identifier; other modes normally use one.

The Actor uses OpenCitations' first-party public REST APIs rather than parsing the human-facing website. An optional operator-side OPENCITATIONS_ACCESS_TOKEN environment secret is sent as an authorization header when configured; it is never exposed as a public input or logged. OpenCitations recommends its free access token for application use, but the public endpoints currently accept bounded anonymous requests.

Unknown but well-formed identifiers can truthfully return no rows. Invalid identifiers are rejected before network work with a correction example. Upstream outages and exhausted rate limits produce warnings and a SUMMARY diagnostic; partial valid results are preserved. The Actor never creates placeholder citation records to make an empty run appear successful.

For bulk harvesting beyond the Actor's 10,000-record run cap, use the official OpenCitations data dumps instead of issuing many API calls. Coverage and counts reflect the OpenCitations indexes at retrieval time and can change as their open datasets are updated.

API use

Call the Actor through the Apify API with the same JSON input used in Console. The default dataset can be exported as JSON, CSV, Excel, XML, or RSS through Apify storage endpoints. AI agents can use the output schema to distinguish citation edges, metadata nodes, and count records without inferring fields from logs.

Legal and responsible use

OpenCitations states that its datasets are released under the CC0 public-domain dedication. The Actor records the exact source endpoint for provenance. Cite OpenCitations when appropriate in research outputs, respect its fair-use limits, and review the source project's current terms and attribution guidance for your use case. This tool does not access paywalled full text, authentication-protected publisher pages, or personal accounts.

For support, include the Apify run ID, selected operation, a non-sensitive example identifier, and the SUMMARY record. Do not include account tokens or other secrets.