Claude Paper plugin
A plugin for studying research papers: a PDF or arXiv link becomes a summary, study materials, code demos and a web viewer
- 353 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- On first run the plugin installs dependencies through npm and pip and starts a local web viewer on port 5815
- The agent downloads PDFs from links and writes files into its own folder in your home directory and into the agents' skill directories
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/claude-paperDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- On first run the plugin installs dependencies through npm and pip and starts a local web viewer on port 5815
- The agent downloads PDFs from links and writes files into its own folder in your home directory and into the agents' skill directories
This is third-party code. Review the repository files before installing.
What it does
Claude Paper turns a research paper into a study folder. You give the agent a PDF path, a direct link or an arXiv page address. The agent parses the file, extracts the metadata and full text, judges the paper's complexity and type and picks a set of materials for it. The full text goes into paper.txt, while the metadata keeps a 50,000-character preview so the context is not overloaded. There are two workflows. The quick one: `/claude-paper:summary` writes a 300–500 word summary for deciding whether a paper deserves a deeper read. The full study produces a README, a detailed summary with key insights and review questions. On top of that the agent writes study implementations with Jupyter notebooks, looks for the original code through GitHub, arXiv and CodeOcean links and pulls out figures and tables. All materials go into a shared library with a search index. `/claude-paper:webui` starts a Nuxt web viewer at http://localhost:5815 with library search, reading with KaTeX math, code and notebooks. The materials follow the language of your request.
Who it is for. For students, graduate researchers and engineers who read AI and machine learning papers and want to grasp the idea fast and try the code.
Good fit when
- You have a dozen papers and need to decide fast which deserve reading
- You want to study one paper in depth with formulas, questions and runnable demo code
- You want to open a paper by an arXiv link and get the PDF and metadata at once
- You want a searchable library of studied papers that you browse in a browser
- You need the same set of materials across Claude Code, Codex, OpenCode and DeepSeek Harness
Not a fit when
- You need to search for papers by topic: the plugin processes files you already picked and does not search literature
- You need a citation library manager: tools such as Zotero cover that
- The source is not a PDF: the plugin takes a PDF path, a direct PDF link or an arXiv page
Example request
Help me study the paper at https://arxiv.org/abs/1706.03762: make the materials, show the code demo and launch the web viewerLimitations
You need Node.js 20.19.x or 22.12 and newer, npm, one of the supported agents and poppler-utils for image extraction. The Claude Code route needs the Claude Code CLI. On first run the plugin installs dependencies and the Python package pymupdf for images.
The model behind your agent writes the study materials, so check the summary and code against the paper.
No MCP server is needed, since the plugin rests entirely on agent skills. Whether arXiv and GitHub are reachable from Russia has not been checked.
How to disable. In Claude Code remove the plugin with `/plugin` and restart the agent. For other agents delete the installed skills from `~/.agents/skills/` and the OpenCode commands. The paper library in `~/claude-papers/` stays on disk and is removed by hand.
FAQ
How do I install Claude Paper in Claude Code?
Run `npx --yes @zlzliqing/claude-paper@latest install --target claude-code`. The installer registers the marketplace and installs the plugin through the official CLI. If you only use Claude Code, the commands `/plugin marketplace add alaliqing/claude-paper` and `/plugin install claude-paper` also work.
Which agents are supported?
Claude Code, Codex, OpenCode and DeepSeek Harness. The install command without `--target` sets up all four. For Codex and DeepSeek Harness the skills go to `~/.agents/skills/`, and OpenCode also gets slash commands.
How does the quick summary differ from the full study?
`/claude-paper:summary` creates only `quick-summary.md` of 300–500 words and keeps the PDF, `paper.txt` and metadata. The full study adds a README, a detailed summary, insights, questions, code and figures and launches the web viewer.
Where are papers stored?
In `~/claude-papers/papers/{slug}/`: the original PDF, full text, metadata, materials, images and code. A global `index.json` for search sits alongside. The folder is created at install if it does not exist yet.
Can I give a link instead of a file?
Yes. A direct PDF link works, and so does an arXiv abstract page address, which the plugin converts to a PDF link itself. For a local file, give its path.
Discussion
No questions or reviews yet.
Related
PluginsTakes an agent from a topic to an arXiv review paper in LaTeX IEEEtran: approved plan, every citation checked against a live source, PDF build
ClaudePrism: LaTeX editor with Claude Code
ClaudePrism
A desktop editor for scientific writing in LaTeX with Claude Code built in and files kept on your disk
A ready-to-fork Claude Code template for academic work: slides, papers and data analysis with multi-agent review and quality gates
Claude Code Docs
Claude Code Documentation Tool
A Claude Code plugin that keeps the official Claude docs at hand, reading each page live from Anthropic on demand, with no mirror and no MCP server
The paper-spine skill takes a topic, materials and data through a scientific paper: literature, text, figures, review and Word, LaTeX and PDF files
A local MCP server for arXiv literature work: section-level paper reads, BibTeX from metadata, topic watches
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Claude Paper (this entry) | no | 353 | ||
| DeepPaperNote | no | 1,269 | a skill for deep-reading a single paper into an Obsidian-style note | |
| arxiv-mcp-server | no | not required | 3,240 | an MCP server for searching and downloading arXiv papers |
README in short
The description opens with a feature list and a quick start: installing for all agents with one command, upgrading and the Claude Code marketplace route. System requirements, usage examples, the paper folder structure and the plugin architecture with adapters for each agent follow.
README badge
Original SKILL.md
--- name: summary description: Use this for a quick summary of a research paper's core ideas and key points. Use when you want to quickly understand a paper without deep study materials. Triggers on PDF paths, arXiv URLs, or paper URLs. disable-model-invocation: false allowed-tools: Bash, Write, Read --- # Quick Paper Summary Workflow This skill generates a **concise summary** of a research paper's core ideas and key points. **When to use:** - You want to quickly understand what a paper is about - You need the main contributions without deep technical details - You're screening papers to decide which to study in depth **When NOT to use:** - You want comprehensive study materials (use `/claude-paper:study` instead) - You need code demonstrations - You want interactive visualizations --- **Language Detection**: Detect the user's language from their input and generate ALL materials in that language. - Example: User says "我们学习一下这篇论文" → Generate materials in Chinese - Example: User says "Let's study this paper" → Generate materials in English --- # Step 0: Check Dependencies (First Run Only)