OpenAlex and Valyu solve similar research problems. Valyu is a search and DeepResearch API built for AI agents, built for retrieving full-text evidence with attached citations across academic literature, scientific research, papers, and specialist sources in a single call. OpenAlex is an open scholarly metadata graph of more than 322 million works, built for bibliometric discovery, citation analysis, and structured queries across works, authors, institutions, and topics.
Quick answer
Valyu gives you usage-based semantic retrieval of full-text passages with structured citations across academic literature, scientific research, papers, journals and specialist sources in a single call for both humans and AI agents. OpenAlex gives you an open scholarly graph of 322 million-plus works with structured metadata, filters, authors, institutions, topics, and citation relationships
What content can OpenAlex index?
OpenAlex indexes scholarly works as a connected graph. A "work" is any scholarly document: a journal article, conference paper, book or book chapter, dataset, dissertation, or preprint.
Works are split across two corpora, which is worth knowing before you compare headline numbers. The core is the curated catalogue and it is what every query returns by default: a live call to api.openalex.org/works in August 2026 gives 322,147,582 works. An expansion layer adds a rawer 165,253,218 records, mostly datasets and single-repository entries, and corpus=all returns both for 487,400,800. Every "320 million" figure you see quoted, including the one in this article, is the core.
The graph connects works to authors, sources (journals, conference proceedings, repositories), institutions, topics, keywords, funders, awards, and publishers. Bibliographic coverage is assembled from upstream indexes including Crossref, PubMed, DataCite, DOAJ, and arXiv.
The search parameter for works matches across title, abstract, and fulltext, and the fulltext.search filter covers the same three fields.
Separately, OpenAlex maintains a content archive of cached full text: 50M+ PDFs (roughly 250 TB) and ~43M TEI XML files parsed by Grobid. Filtering on has_content.pdf:true returns 52,407,611 works, every one of them open access. Retrieval runs through a separate content API at $0.01 per file, which is about 100 files a day on the free $1 budget.
OpenAlex does not provide full text for paywalled or subscription-only publications. It returns metadata and links to locations where full text may be available, and what you can actually retrieve depends on the open-access status of each work.
What content can Valyu index?
Valyu indexes roughly 40 million academic papers alongside web results, financial market data, and licensed and specialist collections. The academic sources break down as:
| Source | Coverage | Update frequency |
|---|---|---|
| PubMed | 37M+ papers | Monthly |
| arXiv | 2.5M+ papers | Monthly |
| bioRxiv | 250K+ papers | Monthly |
| medRxiv | 80K+ papers | Monthly |
| chemRxiv | 30K+ papers | Monthly |
Licensed Wiley collections cover finance journals and finance books (quantitative finance, risk, investment theory) on a monthly cycle, and Wiley Health & Life Sciences journals (medicine, nursing, pharmacology, allied health) quarterly.
Beyond academic content, the index covers clinical trials from ClinicalTrials.gov (500K+, real-time), USPTO patents (8M+, weekly), EPO patents (4M+, weekly), FDA drug labels (150K+, real-time) and others.
Valyu does not provide full text for every indexed publication. Full-text retrieval covers open-access content and sources where access has been licensed, and licensed content carries usage restrictions.
How do the web interfaces compare?
The Valyu platform provides a search playground that queries across source categories and returns title, URL, content, and structured citations. Because retrieval is semantic, an exact-title query matches by meaning rather than string equality, which returns the target work plus related material rather than a single row.
The OpenAlex interface presents a search box and filter options. Boolean operators (AND, OR, NOT) must be capitalised, and quotation marks give you phrase matching. Author searches run through the /authors endpoint, and citation searching runs through citation filters.
Feature comparison
| Feature | OpenAlex | Valyu |
|---|---|---|
| Corpus size | 322M core works, 487M including the expansion layer | ~40M academic papers, plus web, financial, and specialist sources |
| API base URL | api.openalex.org |
api.valyu.ai |
| Authentication | Optional free key ($0.10/day without, $1/day with) |
x-api-key header, required |
| Query type | Keyword, Boolean, phrase, proximity, wildcard, semantic | Semantic full-text retrieval |
| Max query size | ~4 KB URL | Not a documented constraint |
| Filters | Structured entity filters across all endpoints |
search_type, included_sources, excluded_sources, source_biases, relevance_threshold, start_date, end_date, category, country_code, max_num_results, max_price
|
| Pagination |
page / per-page (max 100), 10,000-result ceiling, cursor paging beyond |
max_num_results |
| Response fields | Selectable via select; full metadata schema |
title, url, content, relevance_score, citations |
| Citations | Citation graph: cited_by_count, references, related works |
doi, authors, citation, citation_count, references on academic results |
| Full text | 52.4M works with a cached PDF, all open access | Open-access papers plus licensed collections |
| Web search | Not included | Included |
| Financial data | Not included | 75 exchanges, structured JSON |
| Free tier | $1/day of API usage with a free key | $10 in credits, $20 with a work email |
Freshness, access rights, coverage gaps, and pricing
Freshness. OpenAlex updates continuously from its upstream indexes. Valyu’s indexes daily, monthly and quarterly. If your workflow depends on a paper published last week, Valyu will surface it faster.
Access rights. OpenAlex provides metadata for all indexed works and full text only for open-access ones. Valyu provides full text for open-access papers plus licensed collections where access has been secured. Both operate inside publisher restrictions.
Coverage gaps. OpenAlex’s 322 million core works are bibliographic records, not 322 million retrievable full texts: 52.4 million carry a cached PDF. Valyu’s roughly 40 million academic papers are a fraction of OpenAlex’s record count but gives full-text depth.
Which workflows fit each service?
Valyu fits full-text evidence retrieval and multi-source search:
- AI agent retrieval with structured citations
- Cross-domain research spanning papers, web content, and financial data
- Full-text passage extraction from open-access PubMed and arXiv content
- Specialist retrieval from SEC filings, clinical trials, patents, and regulatory documents
- RAG pipelines needing clean, cited content
- Scientific and academic deepresearch
OpenAlex fits structured metadata discovery and citation graph analysis:
- Bibliometric analysis across authors, institutions, and topics
- Citation tracking and reference mining
- Systematic review screening by publication year, type, or venue
- Large-scale metadata extraction via cursor paging or the bulk snapshot
- Open-access identification and location linking
When do OpenAlex and Valyu work together?
The combined pipeline uses OpenAlex for discovery and Valyu for evidence retrieval:
- Query OpenAlex for works matching a topic, author, or publication window using filters and search parameters.
- Extract DOIs, titles, and metadata from the results. Single-entity lookups by DOI are free, so this stage costs almost nothing.
- Use Valyu to retrieve full-text passages and structured citations for those works or related content.
- Combine OpenAlex's citation graph with Valyu's retrieved passages to build evidence-backed output.
An agent can move across domains in a single run: OpenAlex establishes which works matter and how they connect, Valyu pulls the text that supports a claim. Discovery and retrieval are handled by services designed for each, rather than one tool doing both adequately.
What the same query returns from each in Practice
The two services return different object types for identical input, which is the clearest way to see the design split.
A topic search for "melanoma immunotherapy" on OpenAlex returns work records: titles, citation counts, author and institution links, topic assignments. The same query on Valyu returns full-text passages with citations attached to each one.
The same query on Valyu returns full-text passages with citations attached to each one.
An exact-title search on OpenAlex uses the search parameter or a title filter and resolves to the record. On Valyu, semantic matching returns the target work alongside related content, because it is matching meaning rather than a string.
An author search on OpenAlex uses the dedicated /authors endpoint with affiliation and output data. Valyu has no equivalent author endpoint; author filtering depends on metadata in the indexed source.
A supporting-evidence search on OpenAlex returns related works through the citation graph. Valyu returns passages that address the query directly.
This is a difference in design goal, not in quality. OpenAlex is built for graph discovery. Valyu is built for evidence retrieval.
How do the APIs compare on query syntax and filters?
OpenAlex is a REST API at api.openalex.org. Every entity type is an endpoint: /works, /authors, /sources, /institutions, /topics, /keywords, /funders and more, each supporting list, filter, search, sort, and group operations.
Query syntax supports phrase matching with quotes, proximity search via ~N ("climate change"~5 finds the terms within five words of each other), and unstemmed exact matching plus wildcards through search.exact (machin* matches machine, machines, machinery, with a three-character minimum before the wildcard). Maximum URL length is approximately 4 KB; longer queries return a 400, and the documented fix is to split the Boolean query and combine the returned IDs client-side.
Semantic search is generally available as its own method, priced the same as keyword search, capped at 2,000 characters of input, rate limited to 1 request per second, and returning at most 50 results.
Pagination uses page and per-page, where 100 is the supported maximum (200 is deprecated legacy behaviour). Basic paging reaches only the first 10,000 results, since page * per_page must not exceed 10,000. Cursor paging (cursor=*, then follow next_cursor) goes deeper, and for genuinely bulk work the docs point you at the OpenAlex snapshot instead.
An API key is not required. You can make basic queries with no key at all, at $0.10 of usage per day; a free key raises that 10x to $1 per day. The key travels as ?api_key=KEY or an Authorization: Bearer KEY header, and both work identically.
Valyu is a REST API authenticated with an x-api-key header, with Python, TypeScript, and Rust (alpha) SDKs.
import os
from valyu import Valyu
valyu = Valyu(api_key=os.environ["VALYU_API_KEY"])
response = valyu.search(
"Phase 3 melanoma immunotherapy trials",
search_type="proprietary",
included_sources=["valyu/valyu-pubmed", "valyu/valyu-clinical-trials"],
start_date="2024-01-01",
max_num_results=10,
max_price=20,
)
for result in response.results:
print(result.title, result.url)
print(result.content)
search_type takes four values: all (web plus proprietary, with an LLM router picking sources), web, proprietary, and news. Alongside it the endpoint documents included_sources, excluded_sources, source_biases (soft-rank any source from -5 to +5 instead of filtering it out), max_num_results, max_price, relevance_threshold, start_date, end_date, category, country_code, response_length, and is_tool_call. included_sources accepts dataset ids, bare domains, presets (academic, finance, patent, legal, medical, health, genomics, chemistry, physics, politics, transportation, cybersecurity, compliance, environment, automotive, pulse), and saved collection:<name> groups.
Beyond search, the API surface includes Contents, Answer (grounded answers with citations), and DeepResearch (an autonomous agent with fast, standard, heavy, and max modes producing cited reports and xlsx/docx/pptx/pdf/csv deliverables).
Results carry title, URL, content, source, and relevance_score. Academic results add doi, authors, citation, citation_count, and references, plus abstract and figures where the source provides them. The whole shape is built for LLM context windows.
FAQ
Does OpenAlex require an API key?
No. You can make basic queries with no key at all. A free key raises your daily budget 10x, from $0.10 to $1, and is passed either as ?api_key=KEY or an Authorization: Bearer KEY header.
Does Valyu provide full text for all indexed papers?
Valyu provides full text for open-access papers from PubMed, arXiv, bioRxiv, medRxiv, and chemRxiv, plus licensed Wiley finance and Health & Life Sciences collections. Licensed content carries usage restrictions.
How current is each index?
OpenAlex updates continuously from Crossref, PubMed, DataCite, DOAJ, and arXiv. Valyu refreshes academic sources monthly, SEC filings daily, and clinical trials, drug labels, and market data in real time.
Does Valyu include web search?
Yes. Web search costs $1.50 per 1,000 results. Setting search_type to all queries web and proprietary sources in one call, with a router selecting sources; web, proprietary, and news narrow it.
Can I use both in the same pipeline?
Yes, and free DOI lookups on OpenAlex make it cheap. OpenAlex identifies works by topic, author, or citation relationship, and Valyu retrieves full-text passages and citations for them.
How much does Valyu cost to start?
$10 in free credits with no credit card, or $20 with a work email. Credits work across every source type. Keys come from platform.valyu.ai.
Do preprints count as peer-reviewed papers?
No. Preprints from arXiv, bioRxiv, medRxiv, and chemRxiv are not peer-reviewed, and both services index them alongside peer-reviewed articles. Filter on source or publication type when evidence quality matters.
Top comments (0)