name: arxiv description: Fetch and summarize arXiv papers. Search by topic, read specific papers by ID or URL, and get plain-language summaries. Use when the user mentions arXiv, asks about research papers, wants to find recent academic work on a topic, or is discussing algorithmic or architectural choices that could benefit from literature review. allowed-tools: WebFetch, WebSearch
arXiv
Fetch, search, and summarize academic papers from arXiv.
Usage
/arxiv <query>- Search for recent papers on a topic/arxiv 2401.12345- Fetch a specific paper by ID/arxiv https://arxiv.org/abs/2401.12345- Fetch a paper by URL
Workflow
Search by Topic
- Use the arXiv API to search for papers:
https://export.arxiv.org/api/query?search_query=all:{query}&start=0&max_results=10&sortBy=submittedDate&sortOrder=descending - Parse the Atom XML response to extract: title, authors, abstract, published date, arxiv ID, categories
- Present results in a scannable format
Fetch Specific Paper
When given an arXiv ID (e.g., 2401.12345) or URL:
- Extract the ID from the input
- URL formats:
arxiv.org/abs/XXXX.XXXXX,arxiv.org/pdf/XXXX.XXXXX - Raw ID:
XXXX.XXXXXorXXXX.XXXXXvN
- URL formats:
- Fetch the abstract page:
https://export.arxiv.org/api/query?id_list={id} - Parse and present the paper details
Read Full Paper (HTML)
Many recent papers have HTML versions available:
https://arxiv.org/html/{id}
If the user asks to "read" or "summarize the full paper", try fetching the HTML version first. If unavailable, note that only the abstract is accessible and summarize from that.
Output Formats
Search Results
ARXIV SEARCH: {query}
═══════════════════════════════════════════════════════════════════
1. {Title}
{Authors (first 3, et al. if more)}
{Date} · {Primary category}
{2-3 sentence plain-language summary of the abstract}
https://arxiv.org/abs/{id}
2. {Title}
...
═══════════════════════════════════════════════════════════════════
Single Paper
{Title}
═══════════════════════════════════════════════════════════════════
Authors: {Full author list}
Published: {Date}
Categories: {categories}
URL: https://arxiv.org/abs/{id}
ABSTRACT
───────────────────────────────────────────────────────────────────
{Original abstract}
PLAIN-LANGUAGE SUMMARY
───────────────────────────────────────────────────────────────────
{3-5 sentence summary accessible to a technical but non-specialist reader.
Explain what problem the paper addresses, the approach, and the key results.
Avoid jargon where possible; define it when unavoidable.}
═══════════════════════════════════════════════════════════════════
API Notes
- The arXiv API returns Atom XML. Parse
<entry>elements for results. - Rate limit: be respectful, no rapid-fire queries. One search per invocation is fine.
- The API supports boolean queries:
cat:cs.RO AND all:autonomous(category + keyword) - Common categories:
cs.RO- Roboticscs.AI- Artificial Intelligencecs.LG- Machine Learningcs.CV- Computer Visioncs.CL- Computation and Language (NLP)cs.SE- Software Engineeringstat.ML- Statistics / Machine Learningeess.SY- Systems and Controlmath.OC- Optimization and Control
- For category-specific searches:
search_query=cat:cs.RO+AND+all:{query}
Examples
User: /arxiv behavioral cloning robotics
Claude: [Fetches arXiv API with query "behavioral cloning robotics"]
ARXIV SEARCH: behavioral cloning robotics
═══════════════════════════════════════════════════════════════════
1. Learning to Navigate Sidewalks from Demonstration
J. Smith, A. Lee, et al.
2026-01-15 · cs.RO
Proposes a behavioral cloning approach for sidewalk navigation
using teleoperation demonstrations. Achieves 94% success rate
on unseen environments using a lightweight MLP policy.
https://arxiv.org/abs/2601.09876
...
User: /arxiv 2401.12345
Claude: [Fetches specific paper]
Some Paper Title
═══════════════════════════════════════════════════════════════════
Authors: Jane Doe, John Smith
Published: 2024-01-23
Categories: cs.RO, cs.AI
URL: https://arxiv.org/abs/2401.12345
ABSTRACT
───────────────────────────────────────────────────────────────────
{abstract text}
PLAIN-LANGUAGE SUMMARY
───────────────────────────────────────────────────────────────────
{accessible summary}
═══════════════════════════════════════════════════════════════════