Openalex Search Entities
openalex_search_entitiesRead-onlyIdempotent
Search, filter, sort, or retrieve by ID. Covers all OpenAlex entity types (works, authors, sources, institutions, topics, keywords, publishers, funders). Pass id to retrieve a single entity. Otherwise, use query and/or filters for discovery. Supports keyword search with boolean operators, exact phrase matching, and AI semantic search. Use openalex_resolve_name to resolve names to IDs before filtering. Searches and ID lookups return a curated set of fields by default; pass select to override with specific fields, or ["*"] for the full record. Responses cap at 64,000 bytes per surface unless the least a call can return is larger (over_budget); omitted and windows give the calls that continue a cut, and slice pages a long array.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| id | No | Retrieve a single entity by ID. Supports: OpenAlex ID ("W2741809807"), DOI ("10.1038/nature12373"), ORCID ("0000-0002-1825-0097"), ROR ("https://ror.org/00hx57361"), PMID ("12345678" or "https://pubmed.ncbi.nlm.nih.gov/12345678"), ISSN ("1234-5678"). Keywords are identified by slug rather than a native ID — pass either the slug ("groundwater") or the URL a search returns ("https://openalex.org/keywords/groundwater"). A PMCID is recognized too, bare ("PMC1234567") or as a PubMed Central URL, but OpenAlex indexes no PMCIDs, so it resolves nothing — pass the work's PMID or DOI instead. When provided, `query`, `search_mode`, `filters`, `sort`, `sample`, and `seed` are not applied — the returned record is the entity at that ID regardless of them, and the response `notice` names any you passed. `select` still applies: the curated per-entity-type default is returned unless you pass `select` (use `["*"]` for the complete record). An array too long for the response budget comes back as a window (`windows`); page the rest with `slice`. To filter, drop `id` and search. Use openalex_resolve_name to find the ID if unknown. | |
| page | No | Page number (1-based) for semantic search, the one mode that paginates with `page` instead of `cursor`. Semantic search ranks a query-dependent candidate set whose size `meta.count` reports, so the last reachable page is ceil(meta.count / per_page) — e.g. page 14 for a count of 70 with per_page=5. Passing it under any other search_mode is rejected. | |
| seed | No | Deterministic seed for `sample`. Same seed + same filters = same results — pass when reproducibility matters. Has no effect without `sample`, and a search that passes it alone is rejected. | |
| sort | No | Sort field. Prefix with "-" for descending. Comma-separate for a multi-key sort, applied left to right, with the "-" prefix set per key ("-publication_year,cited_by_count" sorts by year descending, then citations ascending). Common: "cited_by_count", "-publication_date", "-relevance_score" (default when query present). Note: when combined with a keyword query, an explicit sort overrides relevance ranking entirely — top results may be highly cited but only tangentially on-topic. Use "-relevance_score" or omit sort to keep the most relevant results first. "-relevance_score" requires an active search via "query" or a "filter:search" filter — passing it without one will fail. Not combinable with `sample` — a search passing both is rejected. | |
| query | No | Text search query. Supports boolean operators (AND, OR, NOT), quoted phrases ("exact match"), wildcards (machin*), fuzzy matching (machin~1), and proximity ("climate change"~5). Omit for filter-only queries — an empty string is rejected, since a blank search is a mistake rather than a request for the whole catalog. | |
| slice | No | Page one array of the record at `id`: returns `id`, `display_name`, and `field` from `offset` onward, as many elements as fit the response budget (at least one), with the window and its `next` call in `windows`. Requires `id`; replaces `select`. An offset at or past the end returns an empty window. Walks any partial array, including a possibly capped 100-authorship list. | |
| cursor | No | Pagination cursor from a previous response. Pass to get the next page. Omit it on the first call — an empty string is rejected, since a supplied-but-blank cursor is a caller mistake rather than a request for page 1. Keyword and exact modes only — semantic search walks its candidates with `page`, and a `cursor` sent with it is rejected. | |
| sample | No | Return a random sample of this many entities matching the filters (1-100). Single page only — neither `cursor` nor `page` pagination applies to sampling, and a search that passes either alongside it is rejected. Keyword and exact modes only: OpenAlex does not sample a semantic search, so `sample` with search_mode "semantic" is rejected. Cannot be combined with `sort` — a sample has no order, and a search passing both is rejected. Overrides `per_page`. Useful for unbiased exploration: spot-checking filter correctness, stratified review prompts, or generating exploration sets without bias toward most-cited. | |
| select | No | OpenAlex top-level field names to return. Always returned: `id`, `display_name` — additional fields you list are appended. A curated default per entity type applies to both searches and single-entity (`id`) lookups; pass field names to override it, or `["*"]` to retrieve the complete record (every field). Only top-level fields project, so a nested value is requested by its parent object: bibliometrics (`h_index`, `i10_index`, `2yr_mean_citedness`) live under `summary_stats` on authors, sources, institutions, publishers, and funders, and naming a leaf returns that object. Invalid field names produce an error identifying the rejected field. Large projections may be cut to fit the response budget (`omitted`, `windows`). Not applied with `slice`. Example: ["doi", "authorships", "primary_topic"]. | |
| filters | No | Filter criteria as field:value pairs. AND across fields (multiple keys). OR within field: pipe-separate ("us|gb"). NOT: prefix "!" ("!us"). Range: "2020-2024". Comparison: ">100", "<50". AND within same field: "+"-separate. Two keys that resolve to the same upstream field (an alias and its canonical name, e.g. `year` and `publication_year`) are both applied and AND'd, so they narrow rather than override each other. Use OpenAlex IDs (not names) for entity filters — resolve names first. Common keys: `openalex` (filter by entity ID, e.g. {"openalex": "W123|W456"}), `cites` (works citing a given work), `publication_year` (range "2020-2024"), `authorships.author.id`, `type`, `is_oa`. | |
| per_page | No | Results per page (1-100). Default 25. Semantic search caps at 50 — when search_mode="semantic", set per_page ≤ 50 (also subject to a 1 req/sec rate limit upstream). The cap applies to searches only; an `id` lookup returns its one record regardless of both. A budget-cut page returns fewer (`omitted`). | |
| entity_type | Yes | Type of scholarly entity to search. | |
| search_mode | No | Search strategy. "keyword": stemmed full-text (default). "exact": no stemming, matches individual words (use quoted phrases for multi-word exact match). "semantic": AI embedding similarity over a query-dependent candidate set whose size `meta.count` reports, at ~1 req/sec, up to 50 per page, and paginated with `page` rather than `cursor`. | keyword |
Output Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| echo | No | Compact echo of the criteria that actually ran (entity_type, query, filters, sort, search_mode) — surfaces what was searched when results are empty. An `id` lookup echoes entity_type and id alone, because the search criteria are not applied on that path. | |
| meta | No | Result metadata including pagination. | |
| error | No | Present when the call failed. Absent on success. | |
| budget | No | What this call cost against the OpenAlex daily budget and what is left of it. Price a full traversal before committing to it: `totalCount` ÷ `per_page` × `costUsd` against `remainingUsd`. Absent when OpenAlex omitted the accounting headers. | |
| notice | No | Guidance notice. Set when a first call returns no results (echoes the criteria and suggests how to broaden), when a paginated call ran past its last page (says the traversal is finished instead of advising a broader query), when an `id` lookup was passed search criteria it does not apply (names them), on every semantic search to disclose that `meta.count` is a candidate total rather than a match total, when the response budget cut records, windowed an array, or ran over, when a list record carries exactly 100 authorships (possibly capped), when a `slice` offset is past the end of its array, and when `slice` was passed with `select`, which it replaces. Absent otherwise. | |
| omitted | No | Present when the response budget cut whole records from the page end; pagination continues after the full upstream page. | |
| results | No | OpenAlex entity objects. Text values are plain text — HTML entities decoded, HTML/JATS/MathML markup removed — except identifier and URL fields such as `id`, `doi`, `ids`, and `*_url`, returned exactly as OpenAlex stores them; an abstract arrives reconstructed as `abstract`. Additional fields depend on entity_type and select. Records cut by the response budget are in `omitted`; partial or possibly capped arrays are in `windows`. | |
| windows | No | Arrays returned in part — windowed to fit the budget, paged by `slice`, or 100 possibly capped authorships — each with its continuation. A `slice` call always carries its window, even for a whole array; otherwise absent when every array is whole. | |
| totalCount | No | Total results matching the query/filters across all pages. | |
| over_budget | No | True when even the least this call can return — the first record's non-array fields (its arrays windowed to empty), or one `slice` element — exceeds the budget. |