Openalex

Status: active

/openalex -- Academic Paper & Research Search

OpenAlex is a free, open catalog of the global research system (477M+ works, 90M+ authors).

This skill is the lightweight, token-frugal counterpart to the openalex-research-mcp MCP server in this repo: instead of a persistent server with 31 structured tools, it shells out to a small CLI on demand, so it costs nothing while idle. Use the MCP server inside MCP clients (Claude Desktop, TypingMind); use this skill inside coding agents with shell access.

When NOT to use

• arXiv/SSRN full-text search -> OpenAlex has metadata only, not full text
• Reading a specific paper's full text -> fetch the DOI URL directly
• Getting BibTeX citations -> use CrossRef (https://api.crossref.org/works/<DOI>)

Setup

The skill drives the openalex CLI bundled at skill/bin/openalex (zero dependencies, Python 3). Put it on your PATH:

chmod +x skill/bin/openalex
export PATH="$PWD/skill/bin:$PATH"
# Optional but recommended (faster "polite pool"):
export OPENALEX_EMAIL="you@example.com"
# Optional (premium): export OPENALEX_API_KEY="..."

Both env vars are optional — OpenAlex serves anonymous requests for free.

Core Patterns — openalex CLI

The CLI handles key/email injection, mandatory select= field trimming (full responses are 3x larger), abstract reconstruction from the inverted index, and DOI→W-id resolution. Add --json for shaped JSON output.

Search works (papers)

openalex works "supply chain network" -n 10
openalex works "structural estimation" --filter "from_publication_date:2024-01-01,cited_by_count:>10"
openalex works "<query>" --abstract        # include reconstructed abstracts

Get a specific work

openalex work 10.3982/ecta7743             # by DOI
openalex work W2070854196 --abstract --refs

Authors (two-step: search name → use the A-id; names are ambiguous, IDs are not)

openalex authors "daron acemoglu" -n 3
openalex author-works A5012301204 -n 10    # top works by citation
openalex experts "CRISPR gene editing" -n 5  # leading authors ON a topic, by h-index

experts resolves the topic to a topics.id and ranks authors who work on it by h-index (unlike authors, which name-matches) — use it to find field leaders.

Institutions / topics / sources

openalex institutions "seoul national university"
openalex topics "supply chain"             # topics replaced "concepts" — never use concept.id
openalex sources "management science"

Citation network & trends

openalex cites 10.3982/ecta7743 -n 10      # papers citing this work (DOI auto-resolved)
openalex work W1234 --refs                 # references of a paper
openalex trend T10539                      # works per year for a topic

Abstract reconstruction

openalex abstract 10.3982/ecta7743         # inverted index → readable text

Raw API (only when the CLI lacks a filter/field)

Base https://api.openalex.org, append &api_key=$OPENALEX_API_KEY (if set, else &mailto=$OPENALEX_EMAIL) and ALWAYS &select= only the fields you need. Filter syntax, sorting, pagination: see references/api-reference.md.

File Index

Tips

• Two-step for names: openalex authors first to get the A-id, then openalex author-works — names are ambiguous; IDs are not.
• Empty results? Try: broader date range, fewer filters, alternate spelling, or check the topic/author ID via autocomplete.
• The CLI already enforces select= and reconstructs abstracts — only fall back to raw curl for exotic filters.

Output Contract

All queries produce JSON piped through jq for compact display:

• Work search — array of {title, year, cited, doi, venue, authors} objects
• Author search — array of {id, name, works, cited, inst} objects
• Institution/topic/source search — array of {id, name, works, cited} objects
• Citation network — list of OpenAlex IDs or {title, year, cited} objects
• Abstract — reconstructed plain text string

Present results as a markdown table when >=3 items, inline JSON for 1-2 items. Always include the total count from meta.count when available.