What is Nisa

Nisa indexes more than 500,000 arXiv machine-learning papers — the corpus — and lays them out as a navigable map: position by content, colour by community in the citation-and-similarity graph, the highest-PageRank papers visible first and more as you zoom. A research agent works over the same corpus.

Surface	For
Web app — rapidreview.io/nisa	Explore the map, open a paper and its similar papers, search, save and share lists. Web app
Web chat — `⌘K`	Ask the agent: it searches, reads papers, follows citations, runs code, writes reports; share a chat with a colleague. Web chat
`nisa` CLI	Look papers up from a shell, chat with the agent in the terminal, and connect a local folder so the agent can read your PDFs and notes. Install
Coding agents and scripts	Install the Nisa skill and your agent runs the same lookups itself, headless, with an API key. Install → the skill

Browsing and keyword search need no account. Anonymous visitors get five agent searches and ten chat messages; a free account lifts the caps and unlocks lists, sharing, and API keys.

Two kinds of command

Direct free	Chat-backed usage-tracked
Answer from the indexes in about a second: `search`, `semantic-search`, `paper`, `related`, `excerpts`, `cite`, `pdf`, lists	Create a model turn, 30–120 s, rate-limited: `ask`, `exec`, `ask-batch`, `wait`

The Python SDK (rapidreview-sdk) isn’t covered here yet.

Install

One command, then sign in.

First search

Search, open, related, cite.

Web app

The map, lists, settings, shortcuts.

⌘I

Start

Guides

Reference

Two kinds of command

Install

First search

Web app

​Two kinds of command

Install

First search

Web app

Two kinds of command