scholarly-literature-review

Scholarly Literature Review

It is the compliant replacement for any Sci-Hub style MCP.

When to use

• You have a citation, DOI, PMID, arXiv id, or a paper title and want canonical metadata.
• You have a list of identifiers and want a single access-status table across all of them.
• You want to know, per paper, whether the full text is a downloadable open-access PDF, only free-to-read on a publisher or PMC page, behind a subscription, or unavailable.
• You want to explore a paper's references and the works that cite it.
• You want a structured summary built only from legally-available metadata and abstracts.

For a single quick factoid lookup, resolving one identifier may be enough. For broad topical reviews you may layer this with domain-specific ToolUniverse skills.

The MCP tools it uses

Primary tools for this skill (spec dot-name maps 1:1 to the MCP underscore-name):

• paper.batch_resolve -> paper_batch_resolve(identifiers[]) : resolve a list of identifiers into a table of access status, OA links, metadata, plus any unresolved items.
• paper.summarize -> paper_summarize(identifier|doi) : structured digest from legally-available metadata and abstract; the agent writes the prose.
• paper.citation_graph -> paper_citation_graph(identifier|doi, limit) : references and citing works from Semantic Scholar and OpenAlex.

Supporting tools also available on the same server when you need them:

• paper.resolve -> paper_resolve(identifier) : single identifier to canonical metadata, normalized ids (doi, pmid, arxiv, openalex), and confidence.
• paper.fetch_metadata -> paper_fetch_metadata(identifier) : metadata plus abstract and references where legally available.

Access categories (honest)

• subscription : paywalled; cite only.
• unavailable : no available full text exists.
• unknown : access could not be determined.

Workflow steps

1. Gather the identifiers you have: DOIs, PMIDs, arXiv ids, or titles. Titles use a similarity threshold and a too-vague title may come back unresolved.
2. For a set of papers, call paper_batch_resolve to get one consolidated table of metadata, access status, OA links, and unresolved items. For a single paper, use paper_resolve or paper_access_status.
3. Read the access category per paper.
4. To expand a topic, call paper_citation_graph to pull references and citing works, then loop those new identifiers back through paper_batch_resolve.

Concrete examples

• Resolve a landmark methods paper: paper_resolve("10.1371/journal.pmed.0020124") returns the resolved record with title "Why Most Published Research Findings Are False".
• Resolve a preprint by arXiv id: paper_resolve("1706.03762") returns the arXiv record "Attention Is All You Need".
• Classify an open-access paper: paper_access_status(doi="10.1371/journal.pone.0000308") returns oa_pdf with a PLOS OA url.

Limitations

• Title resolution uses a similarity threshold of 0.55; a too-vague title may be reported as unresolved.
• free_to_read_not_redistributable items are intentionally not downloaded; open the landing page instead.
• Citation graph depth depends on Semantic Scholar and OpenAlex coverage for the given DOI.
• An API key that is missing only causes that one source to be skipped, never an error.

Tested invocation

paper_access_status(doi="10.1371/journal.pone.0000308")
-> category: oa_pdf
-> provenance: PLOS open-access url

This is one of the verified live end-to-end checks. The server's test suite covers 22 offline policy/gate tests and 10 live end-to-end checks, and lists 8 tools over MCP stdio.