Claude Research

A complete AI-assisted workflow for scientific articles — from literature review to publication-ready output.

Built entirely inside VS Code with Claude Code. No specialist software, no complex setup — just a folder, an editor, and a conversation.

---

What this is

Claude Research is a template that turns VS Code + Claude Code into a full scientific research environment:

• Automated literature research — Claude searches, evaluates and structures sources with authority, recency and relevance scores
• Interactive citation system — every inline citation (Author, Year) is a live hyperlink; sources are a queryable, filterable database with DOI links, open-access badges and topic clusters
• Article writing — 10-chapter structure in Markdown, APA 7 citations, Austrian/German academic style
• Publication-ready export — build.py generates standalone.html (interactive, printable) and standalone.tex (LaTeX/Overleaf-ready)
• Presentation — Reveal.js slides with AI-generated background images (fal.ai Flux), 3 layout modes, 8 font themes, 18 visual styles
• PowerPoint export — one-click .pptx generation with editable text and embedded background images
• Source download manager — automated PDF download for open-access sources, excerpt extraction for paywalled content
• Course mode — optional syllabus generation with schedule, learning outcomes, required readings and a sample PowerPoint for the first session

---

Screenshots

Source Database

Statistics dashboard — 290 sources, peer-review rate, score distribution, type and language breakdown

Source cards with score badges, citation links and full metadata

Rating methodology — authority, recency and relevance dimensions explained

Article

Article text with clickable inline citations — every (Author, Year) links to the source

Source list with badges, citation status and DOI links

Presentation

Reveal.js slide with AI-generated sticker-style background (fal.ai Flux)

Visual style gallery — 18 styles, one click to switch

---

Requirements

What	Why
VS Code	Editor
Claude Code extension	AI assistant — includes CLI, no Node.js needed
Python 3	For build.py (HTML + LaTeX export)
fal.ai API Key	For AI-generated slide images (optional)

---

Getting started

1. Clone or download this repository
2. Open the folder in VS Code
3. Fill in content/meta.json (title, author, event, sources target, etc.) — or let Claude ask you
4. Start Claude Code (via the extension panel)
5. Type "Leg los" — Claude reads your meta.json and builds everything autonomously

What happens next: Claude researches sources, writes 10 chapters, creates the presentation, verifies all URLs, and hands you the finished article. No questions asked — it works through all 9 steps in one go.

---

API Key setup

Enter your fal.ai key in API Key.js:

window.FAL_KEY = "your-key-here";

Then protect it from accidental commits:

git update-index --skip-worktree "API Key.js"

---

Project structure

claude-research/
├── CLAUDE.md               ← Instructions for Claude Code (the brain)
├── build.py                ← Build script → standalone.html + .tex
├── API Key.js              ← fal.ai key (fill in, then skip-worktree)
├── index.html              ← Landing page with links to all outputs
├── handbuch.html           ← User manual (German)
├── syllabus.html           ← Course syllabus template (optional)
├── content/
│   ├── meta.json           ← Project metadata (title, author, sources target)
│   └── 01–10_*.md          ← Article chapters (Markdown)
├── sources/
│   ├── sources.json        ← All sources with metadata and scores
│   └── pdf/                ← Downloaded source PDFs (auto-populated)
├── assets/css/paper.css    ← Stylesheet
├── presentation/
│   ├── index.html          ← Reveal.js presentation with FAB toolbar
│   ├── stile.html          ← Visual style gallery (18 styles)
│   └── fonts/              ← Local font files (GDPR-compliant)
└── figures/                ← Images for the article

---

Features at a glance

Feature	How
Literature research	Claude searches, evaluates, scores (3–10) and categorizes sources
Interactive citations	Click any (Author, Year) → popup with abstract, DOI, badges
Source quality dashboard	Filter by score, type, badge, language; visual statistics
APA 7 compliance	Inline citations auto-linked, chapter bibliographies generated
URL verification	Every source URL checked for accessibility
PDF download	Open-access PDFs saved locally; paywall → excerpt extraction
HTML article	Responsive, printable, with TOC and citation modals
LaTeX export	Overleaf-ready .tex (color + b/w variants)
Reveal.js presentation	3 layouts, 8 fonts, 18 AI image styles, gradient mode
PowerPoint export	.pptx with editable text and embedded backgrounds
Deployment ZIP	Self-contained package for any web server
Course mode	Syllabus + schedule + sample PowerPoint for first session

---

License

MIT © 2026 Thomas Schroffenegger Brought to life with the rather delightful help of Claude Code (Anthropic)