paper-lookup

name	paper_lookup
description	Use when you need metadata, figures, full text, citations, or BibTeX for a specific paper. Centralizes InspireHEP and arXiv MCP server usage with decision guidelines.

Paper Lookup Skill

Retrieves information about specific papers using two complementary MCP servers: InspireHEP (metadata, figures, citations) and arXiv (full text). Use this skill whenever you need to look up, inspect, or download a paper.

When to Use This Skill

• Getting paper metadata (title, authors, abstract, citation count)
• Downloading figures from a paper
• Reading the full text of a paper
• Getting BibTeX entries for the bibliography
• Checking citation counts or finding citing papers
• Looking up an author's publication list
• Fetching paper references

MCP Server Decision Guide

Need	Use	Tool
Abstract / metadata	InspireHEP	get_paper_details
Full paper text	arXiv	download_paper → read_paper
Figures (with download URLs)	InspireHEP	get_paper_figures
BibTeX entry	InspireHEP	get_bibtex
Citation count / graph	InspireHEP	get_citations
Author publications + h-index	InspireHEP	get_author_papers
References list	InspireHEP	get_references
Non-HEP paper search	arXiv	search_papers
HEP-specific search	InspireHEP	search_papers

arXiv ID Handling

⚠️ Critical: arXiv IDs in XXXX.XXXXX format (e.g., 1012.4515) are parsed as floats by the JSON layer when passed to InspireHEP tools, causing validation errors. This does NOT affect the arXiv MCP server.

MCP Server	Bare ID (1012.4515)	Full URL	Old format (hep-ph/0123456)
arXiv	✅ Works directly	✅	N/A
InspireHEP	❌ Float coercion	✅ Use this	✅ Works directly

Rule for InspireHEP: Always wrap numeric arXiv IDs in the full URL format:

mcp_inspirehep_get_paper_details(arxiv_id="https://arxiv.org/abs/1012.4515")

Old-format IDs with letters work directly:

mcp_inspirehep_get_paper_details(arxiv_id="hep-ph/0512090")

Recipes

Recipe 1: Get Paper Metadata

mcp_inspirehep_get_paper_details(arxiv_id="https://arxiv.org/abs/<ID>")
# Returns: title, authors, abstract, citation_count, DOI, publication info, texkey

Recipe 2: Read Full Paper Text

# Step 1: Download and convert PDF → Markdown
mcp_arxiv-mcp-server_download_paper(paper_id="<ID>")  # bare ID works

# Step 2: Poll until conversion completes (~20-30s for long papers)
mcp_arxiv-mcp-server_download_paper(paper_id="<ID>", check_status=True)

# Step 3: Read the converted markdown
mcp_arxiv-mcp-server_read_paper(paper_id="<ID>")

Papers are stored locally in arxiv-papers/ within the thesis project directory.

Recipe 3: Download Figures from a Paper

Figures are available via InspireHEP for papers that have them indexed.

# Step 1: Get figure URLs and captions
mcp_inspirehep_get_paper_figures(arxiv_id="https://arxiv.org/abs/<ID>")
# Returns: list of {caption, url, description} for each figure

# Step 2: Download a specific figure to the chapter's figures directory
# run_command: curl -L -o "chapter_XX/figures/<bib_key>_figN.<ext>" "<figure_url>"

Save path convention: chapter_XX/figures/<bib_key>_figN.<ext> where:

• <bib_key> is the paper's BibTeX key (e.g., Navarro:1996gj)
• figN is the figure number from the paper
• <ext> is the file extension — check the URL for the format (.png, .pdf, .jpg). If the URL has no extension, try .png first

Important: The InspireHEP MCP server provides figure URLs only — it does not download files. Use curl -L -o or equivalent to download the actual image files from the returned URLs.

The figure URLs point to https://inspirehep.net/files/<hash> and return the original image format (typically PNG or PDF).

Recipe 4: Get BibTeX Entry

mcp_inspirehep_get_bibtex(identifier="<arxiv_id or DOI>")
# For numeric arXiv IDs, use the full URL:
mcp_inspirehep_get_bibtex(identifier="https://arxiv.org/abs/1012.4515")
# Returns: BibTeX entry, texkey, inspire record ID

Recipe 5: Check Citations

# Get paper details first (includes citation_count)
mcp_inspirehep_get_paper_details(arxiv_id="https://arxiv.org/abs/<ID>")

# For the full citation graph (papers that cite this one)
mcp_inspirehep_get_citations(inspire_id="<numeric_inspire_id>", direction="citing")

# For references (papers this one cites)
mcp_inspirehep_get_references(inspire_id="<numeric_inspire_id>")

Recipe 6: Author Publication Search

mcp_inspirehep_get_author_papers(author_name="Cirelli, Marco", sort="mostcited", size=20)
# Returns: paper list + aggregate metrics (h-index, total citations)

Integration with Other Skills

This skill is a utility called by other thesis skills:

Calling Skill	Typical Use
literature_research	Look up abstract/metadata for papers not in NotebookLM; fetch BibTeX for new entries
section_drafting	Download figures for LaTeX inclusion; read paper sections for paraphrasing
review_analysis	Get full text of a review via arXiv when not in NotebookLM
chapter_outline	Check citation impact to prioritize references

Limitations

• InspireHEP is HEP-only — for non-physics papers (cs., stat., etc.), use arXiv search
• arXiv has no figure extraction — PDF→Markdown conversion loses figures
• InspireHEP figures are not universal — not all papers have indexed figures on InspireHEP
• arXiv rate limits — avoid rapid sequential downloads; the server handles one conversion at a time