Command line

kx: KinetixAI in your terminal.

One Python package. Search papers across arXiv, Semantic Scholar and OpenAlex, look up researchers by ORCID, ask grounded questions, pull Paper Lab tabs as text or JSON, query your Research RAG, and browse Primers - all without leaving the terminal. pip install kinetixai and you're in.

Install

pip install kinetixai

Runs on Windows, macOS and Linux. Python 3.9+. search, scholar, get and ask (demo mode) need no account. kx auth login unlocks everything on your plan.

01

Search & metadata

Public APIs. No account needed.

Federated search across three live indexes, de-duplicated and ranked by citations. Fetch metadata, PDFs and BibTeX for any arXiv ID. No login, no rate-limit surprises.

kx search "flash attention" Search arXiv, Semantic Scholar and OpenAlex at once, de-duplicated and ranked.

Example: kx search "sparse mixture of experts" --limit 20 --json | jq '.[].title'

kx scholar "Tri Dao" Look up a researcher by ORCID, Google Scholar link or name and get their recent work.

Example: kx scholar "https://scholar.google.com/citations?user=XXXXXXX"

kx get 2205.14135 --pdf Fetch a paper's full metadata and, optionally, download the PDF.

Example: kx get 1706.03762 --bibtex >> references.bib

kx get 2205.14135 --bibtex Get a paper's citation in BibTeX format, ready to paste into your references.

Example: kx get 2205.14135 --bibtex --json

02

Ask KinetixAI

Demo mode is free. Unlimited use is a Scholar feature.

Grounded, cited answers scoped to a paper or your whole bucket. Same model as the /ai/ page - just in your terminal, pipeable to scripts.

kx ask "What is RLVR?" Ask a grounded, cited question across the demo paper bucket.

Example: kx ask "What is RLVR?" --json | jq '.citations'

kx ask "What is RLVR?" --paper tricalrag Scope the question to a single paper in your bucket by its slug or arXiv ID.

Example: kx ask "What attention mechanism does this use?" --paper 1706.03762

03

Paper Lab

1 paper/day on Explorer, unlimited on Scholar and Professional.

Every Paper Lab tab - architecture, mnemonic, cheatsheet, flashcards, visualization - available as plain text or structured JSON. Pipe into a script, save to a file, or feed into your own tools.

kx paper tricalrag --tab architecture Pull one tab as text. Tabs: visualization, architecture, mnemonic, cheatsheet, flashcards.

Example: kx paper tricalrag --tab architecture --json > arch.json

kx paper 1706.03762 --tab flashcards Get a paper's flashcards for drilling, by arXiv ID or Paper Lab slug.

Example: kx paper 1706.03762 --tab flashcards --json | jq '.[].front'

kx paper 2205.14135 --tab cheatsheet Pull the cheatsheet tab - equations, key facts, notation - as structured text.

Example: kx paper 2205.14135 --tab cheatsheet > cheatsheet.md

04

Your Research RAG

Scholar and Professional, your private workspace.

Add PDFs, notes, arXiv links or GitHub repo URLs to your private Research RAG and query across all of them with the same cited-answer model. Every result traces back to the exact page or line it came from. Papers in your Paper Lab bucket are already in your RAG with no upload needed. Researcher catalogs added via kx scholar --add-rag sync automatically as new papers are published.

kx rag add paper.pdf Add a PDF, arXiv ID, plain-text note, or GitHub repo URL to your private Research RAG.

Example: kx rag add ./papers/ --recursive

kx rag ask "How did we handle fault injection?" Ask across only your own files, cited back to the exact page or line.

Example: kx rag ask "What did we try for the embedding layer?" --json | jq '.answer'

kx rag list List every document currently in your Research RAG.

Example: kx rag list --json | jq '.[].name'

kx scholar "Tri Dao" --add-rag Look up a researcher by name or ORCID and add all their public papers to your RAG at once.

Example: kx scholar 0000-0002-6480-1955 --add-rag --json

05

Primers

Browse free. Buying or downloading needs an account.

Browse the full Primers catalog from the terminal, filtered by category or keyword. Download any Primer you've already bought as a PDF.

kx primers list --category gpu Browse the Primers catalog filtered by category (gpu, ml, systems, math, quantum).

Example: kx primers list --json | jq '.[] | select(.category=="ml") | .title'

kx primers get paper-tricalrag --pdf Download a Primer you already own as a PDF.

Example: kx primers get paper-tricalrag --pdf --out ~/Downloads/

06

Account

Sign in once. Every command follows your plan from that point.

Sign in once. Every command follows your plan from that point: Scholar or Professional unlocks unlimited ask, your Research RAG, and full Paper Lab access.

kx auth login Sign in once; every command above then follows your plan and reaches your own Research RAG.

Example: kx auth status

kx auth status Check which account is signed in and which plan it's on.

Example: kx auth logout

Global flags

These flags work across all kx commands. Combine them freely.

--json

Output structured JSON instead of formatted text. Works on every command.

--limit N

Return at most N results. Applies to search, rag list, primers list.

--out PATH

Write output to a file instead of stdout.

--paper SLUG

Scope kx ask to one paper by slug or arXiv ID.

--tab TAB

Choose a Paper Lab tab (architecture, mnemonic, cheatsheet, flashcards, visualization).

--pdf

Download the PDF when fetching paper metadata.

--bibtex

Output BibTeX citation when fetching paper metadata.

--recursive

Recursively add all PDFs in a directory to Research RAG.

Who uses kx, and how

Researchers

You have a folder of PDFs and need answers from all of them, not just one.

kx rag add ./papers/ --recursive then kx rag ask anything, cited back to the page.

Students

You're drilling flashcards the night before an exam.

kx paper 1706.03762 --tab flashcards gives you a JSON array. Pipe it into any flashcard tool.

Engineers

You want benchmark numbers from a paper in a CI script.

--json on every command. kx ask --json | jq '.answer' and you're done.

Note-takers

You read papers in the terminal and want BibTeX ready without switching apps.

kx get 2205.14135 --bibtex >> references.bib in one step.

Teams

You want a shared Research RAG your whole team can query.

Everyone on a Professional plan runs kx auth login and kx rag add to the same workspace.

Automation builders

You want to automate paper triage, extraction or summarization in a pipeline.

Every kx command outputs clean JSON. Glue it into any pipeline with jq or Python.

Honest about where this is

What works today, and what's in progress

Works right now

Search across arXiv, Semantic Scholar and OpenAlex in one call
Grounded ask on demo papers, free, no account
Full Paper Lab tab access (Scholar and Professional)
Private Research RAG, add and query your own PDFs, notes and repos
Primers catalog browse and download
Structured JSON output on every command
BibTeX export for any arXiv paper

Coming next

kx search --since for date-filtered feeds
kx rag share to share a RAG workspace with a team
kx ask --cross-paper for multi-document reasoning across two or more papers
kx rag catalog for browsing and querying the public RAG catalog from the terminal
Shell completion for bash, zsh and fish
GitHub Actions native example in the docs

Full CLI access - unlimited ask, Research RAG and Paper Lab - is a Professional feature. search, scholar, get and kx ask in demo mode are always free with no account.

Common questions

Do I need an account to use kx? No. kx search, kx scholar, kx get and kx ask (demo mode) all work without an account. kx auth login unlocks your plan's limits for ask, Paper Lab and Research RAG.

Does kx work on Windows? Yes. pip install kinetixai works on Windows, macOS and Linux with Python 3.9+.

Can I use kx in a script or CI pipeline? That's the point. Add --json to any command and pipe the output to jq or your own code. Every command writes clean JSON with no extra formatting when --json is set.

How is kx ask different from asking ChatGPT about a paper? kx ask is grounded in the paper's extracted content - architecture, results, equations - and every answer cites which tab it came from. No training-data guessing.

What's the difference between kx ask and kx rag ask? kx ask queries the KinetixAI demo paper bucket (or one paper in it with --paper). kx rag ask queries only your own private Research RAG - files you added yourself.

Can I add any PDF to my Research RAG? Yes. kx rag add accepts any PDF, plain-text file, or GitHub repo URL. Your files are private and never used to train our models.

Does kx replace the web interface? No - it gives you the same capabilities in a scriptable form. 3D visualizations and the Paper Lab UI still live in the browser.