Source	Method	Best For
arXiv API	WebFetch	Quick metadata lookup, search
arxiv.org	Actionbook + agent-browser	Latest papers, trending, advanced search
ar5iv.org	Actionbook + agent-browser	Section-level reading, figures, citations

┌─────────────────────────────────────────────────────────────┐
│                     arxiv-viewer                            │
├─────────────────┬─────────────────┬─────────────────────────┤
│   arXiv API     │  arxiv.org Web  │      ar5iv.org          │
│   (WebFetch)    │  (Actionbook)   │    (Actionbook)         │
├─────────────────┼─────────────────┼─────────────────────────┤
│ • Metadata      │ • Latest list   │ • Read sections         │
│ • Search        │ • Trending      │ • Extract figures       │
│ • By ID lookup  │ • Advanced      │ • Extract citations     │
│                 │   search form   │ • Paper outline         │
└─────────────────┴─────────────────┴─────────────────────────┘

Key Features

1. API-Based Access (Fast & Structured)

Paper lookup - Get metadata, abstract, authors by arXiv ID
Search - Query papers with field prefixes and boolean operators
Download - Download PDFs directly

2. Web-Based Access (Actionbook + agent-browser)

Latest papers - Browse recent submissions by category
Trending - See featured papers from arXiv homepage
Advanced search - Use arXiv's advanced search form

3. HTML Paper Reading (ar5iv.org)

Section reading - Read specific sections (Introduction, Methods, etc.)
Paper outline - Get table of contents with section IDs
Figure extraction - Extract all figures with captions
Citation extraction - Get full bibliography

Installation

From Actionbook Marketplace (Recommended)

# Step 1: Add Actionbook marketplace
/plugin marketplace add actionbook/actionbook

# Step 2: Install arxiv-viewer plugin
/plugin install arxiv-viewer@actionbook-marketplace

Manual Installation

Clone the repository and add to your Claude Code configuration:

git clone https://github.com/actionbook/actionbook.git
cd actionbook/playground/arxiv-viewer
./setup.sh

Setup

Quick Setup

./setup.sh

Manual Setup

Add to .claude/settings.local.json:

{
  "permissions": {
    "allow": [
      "Bash(curl *)",
      "Bash(mkdir -p *)",
      "Bash(agent-browser *)"
    ]
  }
}

Commands

API Commands

Command	Description	Example
`/arxiv-viewer:paper <id>`	View paper metadata	`/arxiv-viewer:paper 2301.07041`
`/arxiv-viewer:search <query>`	Search papers	`/arxiv-viewer:search transformer --category cs.CL`
`/arxiv-viewer:download <id> [dir]`	Download PDF	`/arxiv-viewer:download 2301.07041 ./papers`

Web Commands (Actionbook)

Command	Description	Example
`/arxiv-viewer:latest <category>`	Latest submissions	`/arxiv-viewer:latest cs.CL`
`/arxiv-viewer:trending`	Trending papers	`/arxiv-viewer:trending`

HTML Reading Commands (ar5iv)

Command	Description	Example
`/arxiv-viewer:read <id> [section]`	Read paper section	`/arxiv-viewer:read 2301.07041 methods`
`/arxiv-viewer:outline <id>`	Get paper outline	`/arxiv-viewer:outline 2301.07041`
`/arxiv-viewer:figures <id>`	Extract figures	`/arxiv-viewer:figures 2301.07041`
`/arxiv-viewer:citations <id>`	Extract bibliography	`/arxiv-viewer:citations 2301.07041`

Report Generation

Command	Description	Example
`/arxiv-viewer:report <id>`	Generate AI paper report	`/arxiv-viewer:report 2301.07041`

Usage Examples

Basic Paper Lookup

/arxiv-viewer:paper 1706.03762

Returns title, authors, abstract, categories, and links.

Search with Filters

# Search by title
/arxiv-viewer:search ti:transformer --max 10

# Search by author
/arxiv-viewer:search au:hinton

# Combined search
/arxiv-viewer:search ti:attention AND cat:cs.CL --max 20

Browse Latest Papers

# Latest NLP papers
/arxiv-viewer:latest cs.CL

# Latest ML papers with limit
/arxiv-viewer:latest cs.LG --max 20

Read Paper Sections (ar5iv)

# Get paper structure first
/arxiv-viewer:outline 2301.07041

# Read specific section
/arxiv-viewer:read 2301.07041 introduction
/arxiv-viewer:read 2301.07041 #S3

# Extract all figures
/arxiv-viewer:figures 2301.07041

# Get bibliography
/arxiv-viewer:citations 2301.07041

Generate Paper Report

# Generate a comprehensive report with AI analysis
/arxiv-viewer:report 2301.07041

# Search by title and generate report
/arxiv-viewer:report "Attention Is All You Need"

The report includes:

Paper metadata (authors, affiliations, date)
Abstract summary
Problem statement
Key contributions
Method overview
Experimental results
Significance analysis
All branded with Powered by ActionBook

Natural Language

"Show me the latest NLP papers"
"Search for transformer papers by Vaswani"
"Read the methods section of paper 2301.07041"
"What figures are in the attention paper?"
"List all citations from this paper"
"Generate a report for paper 2301.07041"

Search Query Syntax

Field Prefixes

Prefix	Field	Example
`ti:`	Title	`ti:transformer`
`au:`	Author	`au:hinton`
`abs:`	Abstract	`abs:attention`
`cat:`	Category	`cat:cs.AI`
`all:`	All fields	`all:neural network`

Boolean Operators

Operator	Usage	Example
`AND`	Both terms	`ti:transformer AND cat:cs.CL`
`OR`	Either term	`au:hinton OR au:lecun`
`ANDNOT`	Exclude	`ti:attention ANDNOT ti:self`

Common Categories

Code	Field
`cs.AI`	Artificial Intelligence
`cs.CL`	Computation and Language (NLP)
`cs.CV`	Computer Vision
`cs.LG`	Machine Learning
`cs.NE`	Neural and Evolutionary Computing
`cs.SE`	Software Engineering
`stat.ML`	Statistical Machine Learning

ar5iv HTML Reading

ar5iv.org converts arXiv LaTeX papers to accessible HTML5, enabling precise section-level access.

Benefits

Feature	PDF	ar5iv HTML
Section access	Full document only	Specific sections
Figure extraction	Difficult	Easy with selectors
Citation parsing	Complex	Structured
Token efficiency	Low (full text)	High (targeted)

Section Selectors

Section	Selector
Abstract	`.ltx_abstract`
Introduction	`#S1`
Methods	`#S3` (varies)
Conclusion	Last section
Bibliography	`.ltx_bibliography`

Workflow

1. /arxiv-viewer:outline 2301.07041    # See structure
2. /arxiv-viewer:read 2301.07041 #S3   # Read specific section
3. /arxiv-viewer:figures 2301.07041    # Get figures

Architecture

Agents

Agent	Model	Purpose
paper-fetcher	haiku	Fetch paper metadata via API
search-executor	haiku	Execute API search queries
paper-summarizer	sonnet	Download and summarize PDFs
browser-fetcher	haiku	Scrape arxiv.org web pages
html-reader	haiku	Read ar5iv.org HTML papers

Data Flow

User Request
    │
    ├─► API Route (paper, search, download)
    │       └─► WebFetch → arXiv API → Response
    │
    ├─► Web Route (latest, trending)
    │       └─► Actionbook → agent-browser → arxiv.org → Response
    │
    └─► HTML Route (read, outline, figures, citations)
            └─► Actionbook → agent-browser → ar5iv.org → Response

Dependencies

Dependency	Purpose
Actionbook MCP	Web page selectors
agent-browser	Browser automation
curl	PDF downloads

Troubleshooting

"No HTML version available"

Not all papers have ar5iv versions. Fallback to PDF:

/arxiv-viewer:paper 2301.07041
/arxiv-viewer:download 2301.07041

"agent-browser not found"

Install agent-browser globally:

npm install -g agent-browser

Permission denied

Run setup script or add permissions manually:

./setup.sh

License

MIT