Q: What can I do with this that I can't do with PDFs?

Query multiple papers like a database. Get all equations, figures, or theorems in one API fetch. Jump to "Section 3.2" by ID. Feed figure URLs straight to vision models. Inject clean Markdown into LLM context without preprocessing. PDFs require complex and lossy extraction pipelines. ScienceStack returns structured data with math, sections, and figures as first-class objects — ready for LLMs and programmatic use. Features include: parsed from LaTeX source, pre-parsed low-latency responses, stable node IDs, clean LaTeX equations, AI summaries, figure URLs, structured tables, Markdown/LaTeX/text export, resolved citations, and section hierarchy.

Q: Why not just use LLMs or OCR to extract from PDFs?

LLMs can read PDFs, but they hallucinate equation details, miss subscripts, and can't reliably extract structured data like node IDs or figure URLs. Every call costs tokens and latency, and long papers hit output token limits. OCR struggles with math — common issues include confusing "k times T" with "k transpose" and misinterpreting square root expressions. PDF extraction pipelines are lossy and complex. PDFs have no semantic structure — no "this is Section 3" or "this is Equation 7". You get coordinates and glyphs, not an AST. Figures are embedded bitmaps with no clean URLs. Tables are just positioned text. ScienceStack parses the LaTeX source directly. You get a structured AST with stable IDs and clean LaTeX — pre-computed, deterministic, and ready to use.

Q: How is this optimized for agentic workflows?

Single-call context: /overview returns TOC, summaries, and lists of all figures, tables, and equations in one request. Stable IDs: Reference sec:3.2 or eq:7 across sessions. IDs don't change, so you can cache and reuse them. Vision-ready figures: Direct image URLs in optimized webp format. Feed straight to vision models. Multiple formats: Get content as JSON AST, Markdown, plain text, or LaTeX depending on your use case. Exact math from source: Equations are extracted directly from LaTeX, not OCR. No broken symbols or missing subscripts.

Q: How do I know if the output matches the source correctly?

Every paper at sciencestack.ai/paper/{arxivId} is rendered using the exact same data the API returns. What you see on the site is what you get from the API. You can visually verify any paper's parsed output before using it programmatically.

Q: What papers are available through the API?

All papers in sciencestack.ai/explore are fully rendered via the API. We currently have 300k+ arXiv papers indexed, primarily in CS, Math, and Physics. Can't find an arXiv paper? Go to sciencestack.ai/paper/{arxivId} and click the "Request" button — we'll parse it within minutes. Alternatively, you can upload LaTeX files yourself and access them via the API, using the paper ID from the URL.

Q: When should I NOT use ScienceStack?

If you just need high-level summaries or general Q&A about papers, feeding PDFs directly to an LLM is much simpler. ScienceStack is built for precision — exact equations, specific sections, figure URLs, stable IDs for caching, or structured data you can query programmatically. While we do support PDF uploads (1 AI credit per page), LaTeX source provides significantly better quality with preserved semantic structure.

Q: How often do you index new papers?

We index papers daily, prioritizing the latest daily submissions from arXiv. We primarily index CS, Math, and Physics papers. Most papers available through the API are from October 2025 onwards. Can't find a paper? Go to sciencestack.ai/paper/{arxivId} and click "Request" — we'll parse it within minutes.

Q: Why do some papers not have AI summaries?

AI summaries are generated asynchronously after parsing completes. For newly indexed or requested papers, there may be a short delay before summaries become available. Occasionally, summary generation may fail for certain papers. If a paper is missing summaries, check back in a few minutes — they're typically generated shortly after the paper is parsed.

Q: How do I upload my own papers?

You can upload your papers through My Library → Uploads. We support both LaTeX and PDF uploads. LaTeX (recommended) is free to process and preserves semantic structure, cross-references, citations, and numbering. PDF uses 1 AI credit per page and uses AI-based extraction.

Q: Can I access my own papers via API?

Yes. Your API key can access both public papers and your own uploaded papers. Uploaded papers count toward your monthly quota the same as public papers.

Question 1

Can I use ScienceStack for free?

Accepted Answer

Yes! The research platform will always be free. You can browse and read all arXiv papers with full features like hover citations, equation previews, and dependency graphs at no cost. The Plus plan unlocks additional features like increased upload limits, larger file sizes, and asset exports (markdown/LaTeX with figures included).

Question 2

How is this different from arXiv HTML or other HTML viewers?

Accepted Answer

Other HTML renderers (including arXiv's LaTeXML output) are fine for short papers, but on long or equation-heavy documents they can become sluggish, with pages that lag, freeze, or even crash the browser. Interactivity is minimal — no hover math previews, no one-click LaTeX copy, no dependency graphs. ScienceStack was built to solve those gaps. Everything is compiled into structured JSON first, then rendered through a virtualized view so even very large math or physics papers stay responsive. All blocks, equations and references are preserved with stable identifiers, which makes hover previews, exact LaTeX copy, fullscreen figures, and automated dependency graphs possible. Every paper also comes with downloadable markdown and JSON exports, so you can take the structured content into your own tools and workflows.

Question 3

Can I upload my own papers? Are uploads private?

Accepted Answer

Yes, you can upload your papers through 'My Library' > Uploads. We support both LaTeX and PDF uploads. LaTeX (recommended) is free to process and preserves semantic structure, cross-references, citations, and numbering. PDF uses 1 AI credit per page and uses AI-based extraction. Uploads are private by default, with the option to share or make public if you choose. We do not index or expose private uploads.

Question 4

How does your markdown export compare to pandoc or other LaTeX→Markdown tools?

Accepted Answer

Most generic LaTeX→Markdown converters (like pandoc) are designed as general-purpose converters, not as faithful scientific parsers. They often drop section numbers, lose equation numbers, break on custom packages, and are not as robust for real-world LaTeX. Our parser was purpose-built for research papers and is much more robust: every section, equation, table, figure, and theorem keeps its original number; all cross-references are preserved with their numbers and converted into live links; it handles many more edge cases and LaTeX primitives than Pandoc; the output is structured enough that you can reliably ask an LLM to "explain equation (3.12)" or "summarize Section 4.1" — and it will have the right context.

Question 5

Which download format is best for AI/LLMs?

Accepted Answer

Both Markdown (.md) and JSON (.json) work great for LLM ingestion. Markdown is best for direct LLM use and human readability — it's lightweight, token-efficient, and every section/equation/table/figure is numbered with preserved cross-references. JSON is best for structured pipelines and tooling — it provides a clean, machine-readable hierarchy with stable IDs for every block, ideal for RAG pipelines, chunking, or filtering. Both formats preserve numbers and references with proper IDs, reducing the risk of LLM hallucination across large paper contexts. Use JSON if you want the LLM to generate new content (e.g. LaTeX) based on the paper.

Question 6

Will this integrate with note-taking or reference managers?

Accepted Answer

Every paper ships with markdown (.md) downloads by default, allowing you to export papers as clean markdown files for use in your preferred note-taking tools. The markdown export works well with VSCode, GitHub MD, Obsidian, and Notion. Our markdown format adds semantic info like equation numbers, section numbers, and table/figure/theorem numbers, and auto-generates linkable references across the file. Direct integration into tools like Obsidian or Notion is on our roadmap.

Question 7

Is ScienceStack accessible and WCAG compliant?

Accepted Answer

Yes, ScienceStack is built to be accessible and compliant with WCAG 2.1 AA standards. Mathematical content includes proper alternative text for screen readers, interactive elements are keyboard navigable, and we maintain proper focus management throughout the interface. This ensures that researchers using assistive technologies can fully access and navigate academic papers.

Question 8

What are AI credits and how do they work?

Accepted Answer

AI credits let you chat with papers using our AI assistant. Each message costs 1-5 credits depending on the model you choose (faster models cost less). Free users get 30 credits/month. Plus users get 300 credits/month. Credits refresh automatically 30 days after your last refresh. You can check your credit balance and next refresh date in your profile settings.

Question 9

What happens to my uploads/annotations if I downgrade from Plus to Free?

Accepted Answer

Your existing uploads remain accessible, but you won't be able to upload new papers beyond the free tier limits until you upgrade again. Your annotations are never deleted. Your AI credits will reset to 30/month on your next refresh date.

Question 10

Can I cancel anytime?

Accepted Answer

Yes, you can cancel your subscription at any time. There are no long-term contracts or cancellation fees. Your account will remain active until the end of your current billing period.

Question 11

Do you offer student or academic discounts?

Accepted Answer

We're currently offering early adopter pricing for all users. The research platform is free forever, and Plus features start at just $5/month — designed to be affordable for students and researchers. You can write to support@sciencestack.ai for any other questions.

ScienceStack Research Pricing

Free

Plus

Need more AI credits?

Compare plans

Free

Plus

AI

Research Tools

Exports

Uploads

Annotations

Frequently Asked Questions