cyanheads OpenAlex MCP Server

Quick Verdict

cyanheads OpenAlex MCP Server is a community-maintained, Apache 2.0-licensed MCP integration for OpenAlex scholarly metadata. Version 0.7.4 provides entity search, trend aggregation, name resolution, one-hop citation exploration, and field discovery through STDIO or Streamable HTTP. Its unified search covers all eight OpenAlex entity types: Works, Authors, Sources, Institutions, Topics, Keywords, Publishers, and Funders. It is useful for literature discovery, research-landscape mapping, entity disambiguation, and expanding a citation neighborhood. It is not a full-text library, a peer-review engine, or an authority that guarantees every returned record is correct.

For local npm execution, plan on Node.js 24 or newer. An OpenAlex API key and a mailto parameter are not interchangeable. OPENALEX_API_KEY is an account credential tied to current authentication, quota, budget, or usage-pricing rules. OPENALEX_MAILTO is only a contact identifier sent to OpenAlex; an email address does not replace a key. The community also provides a public hosted endpoint. That is convenient for low-risk evaluation, but the operator does not control its authentication, access logs, capacity, upgrade timing, or availability. Sensitive or reproducible research should use a pinned self-hosted version.

Best For

• Researchers and students using an AI assistant for initial literature discovery, author disambiguation, and topic exploration.
• Libraries and research-information teams querying open scholarly metadata across several entity types with human verification.
• MCP application developers needing a research tool layer that combines entity search, aggregation, and citation-neighborhood access.
• Literature-review support workflows that need candidate generation and query expansion while retaining formal search and screening records elsewhere.
• Not ideal for users requiring subscription full text, authoritative peer-review conclusions, guaranteed citation formatting, deep graph analytics, or an enterprise SLA from the public instance.

Key Features

• Search across eight entity types: Works, Authors, Sources, Institutions, Topics, Keywords, Publishers, and Funders share one discovery interface.
• Identifier handling: queries can work with OpenAlex identifiers and supported external identifiers such as DOI, ORCID, ROR, PMID, PMCID, and ISSN.
• Trend and distribution analysis: group supported records by publication year, open-access status, institution, country, topic, or other valid fields.
• Name resolution: autocomplete can turn a partial human-readable name into an OpenAlex ID before filtering, reducing ambiguity for authors and institutions.
• One-hop citation graph: explore citation, cited-by, or related-work neighbors around a seed Work without manually constructing OpenAlex filters.
• Field discovery: ask which filter, group, and selection fields are valid before composing a query, reducing upstream parameter errors.
• Two transports and several deployment paths: version 0.7.4 supports local STDIO, HTTP, Docker, and a community-hosted endpoint.

Use Cases

Topic discovery is the most direct use case. Resolve ambiguous names, search Works, and narrow the population by date, open-access state, institution, source, or citation-related fields. For landscape analysis, aggregate a topic by year, institution, source, author, country, or funder, then export the result and validate representative records. The aggregation is a map of the catalog, not an explanation of causality or research quality.

Name resolution helps with authors and institutions that have aliases, spelling variants, or shared names. After obtaining an ID, inspect works, affiliations, external identifiers, and time ranges rather than trusting the first autocomplete result. Citation-neighborhood exploration starts from a known paper and retrieves one-hop citation, cited-by, or related-work candidates for later reading. An agent can also call field discovery before constructing filters, which reduces invented parameter names. Human review is still needed because syntactically valid queries can embody a poor research question.

Pricing

“Freemium” reflects the complete dependency chain more accurately than “free.” The MCP server source is free, while upstream OpenAlex account, quota, and usage-pricing rules should be checked at the time of deployment. OPENALEX_MAILTO can identify a courteous caller but does not confer API-key authentication or budget. Pin version 0.7.4 for production and keep regression queries for upstream field or response changes.

Pros

• One search tool spans all eight core OpenAlex entity types rather than stopping at papers and authors.
• Trend aggregation, name resolution, field discovery, and one-hop citation exploration form a coherent research-discovery toolkit.
• Apache 2.0 licensing plus STDIO and HTTP modes support both local clients and service deployment.
• Separate API-key and mailto settings reflect the distinction between account authority and caller identification.
• Version 0.7.4 provides a concrete baseline for pinned client configuration and regression examples.

Cons

• OpenAlex primarily provides scholarly metadata and selected abstract information, not guaranteed full text for every Work.
• Aggregated records can contain duplicates, omissions, delayed citations, mistaken merges, and uneven disciplinary or language coverage. A model may present these errors with false confidence.
• Citation exploration is one hop around a seed Work, not a deep graph-traversal, path-analysis, or complete bibliometrics platform.
• Users do not control the public instance’s logs, authentication, capacity, upgrades, or availability.
• API keys remain subject to upstream account and budget rules. A mailto value does not replace a key and should not be fabricated.
• Node.js 24+ may be newer than an existing MCP client environment, requiring a separate runtime or upgrade.

Alternatives

FAQ

What is the difference between an OpenAlex API key and mailto?

OPENALEX_API_KEY is a secret generated for an OpenAlex account and participates in current authentication, quota, budget, or pricing behavior. OPENALEX_MAILTO is a contact email parameter used to identify the caller. An email is not an API key, and the key should not appear in shared client configuration or logs.

Does version 0.7.4 cover every OpenAlex entity type?

Its unified search covers Works, Authors, Sources, Institutions, Topics, Keywords, Publishers, and Funders. Valid filter, grouping, and selection fields differ by type, so field discovery should precede unfamiliar queries.

Can its citation graph perform multi-hop network analysis?

The built-in tool explores one-hop citation, cited-by, or related-work neighbors around a seed Work. Export controlled results to a dedicated graph-analysis tool for multi-hop traversal, paths, centrality, or larger network studies.

Is the public hosted instance suitable for production?

It is appropriate for quick, non-sensitive evaluation. Production adoption requires decisions about authentication, logs, limits, availability, upgrades, and data handling. If those boundaries have no formal commitment, self-host pinned 0.7.4.

Can returned metadata be inserted directly into a paper’s references?

Not without checking. Verify title, authors, year, version, venue, and retraction status against the DOI record, publisher, author page, or original publication. Preprints and published versions can otherwise appear as misleading duplicates.

Bottom Line

cyanheads OpenAlex MCP Server 0.7.4 gives an agent a useful interface across eight scholarly entity types, trend analysis, name resolution, field discovery, and one-hop citation neighborhoods. Its role is to broaden and organize research leads, not replace full-text databases, systematic-review methods, or manual fact checking. Use the public endpoint only for non-sensitive trials. For real research, self-host a pinned version, use a dedicated API key, retain queries and screening decisions, and sample-check DOI records and original publications. That process turns broad metadata coverage into traceable evidence rather than polished but unreliable answers.