Question 1

Do you store my data?

Accepted Answer

For public sources, we store embeddings and chunks on our servers. For private repositories, files, and local folders, everything stays on your machine. You can also self-host the entire stack for full data sovereignty.

Question 2

Does Nia support end-to-end encryption for personal data?

Accepted Answer

Yes. Nia supports zero-knowledge end-to-end encrypted sync for sensitive personal data including iMessage, WhatsApp, Apple Notes, Contacts, Reminders, and screenshots. Encryption, embedding, and decryption all happen on your device. Nia's servers only ever see ciphertext and vectors, never plaintext.

Question 3

How is Nia different from a vector DB or Elasticsearch?

Accepted Answer

Vector DBs and Elasticsearch are storage primitives — you still build the chunker, embedder, reranker, and retrieval logic on top. Nia is the retrieval system itself: ingest any source, give your agent a tool to traverse and search it, and get back answers with citations. No RAG pipeline to maintain.

Question 4

Can I self-host Nia?

Accepted Answer

Yes. Nia ships as a self-hostable stack — run the indexer, vector store, and API on your own infrastructure. See the Self-Hosting page for deployment guides.

Question 5

What types of content can I index?

Accepted Answer

Code (GitHub repos, npm/pip/cargo packages), documentation sites, PDFs, research papers, books, mechanical drawings, HuggingFace datasets, local folders, Google Drive, Slack, X, and any public URL. Document Agent and Tracer handle long-form documents with page-level citations.

Question 6

How does Nia handle PDFs, papers, and other long documents?

Accepted Answer

Document Agent traverses long documents the way a human would — table of contents, sections, cross-references — and returns answers with page-level citations. Works on research papers, books, contracts, technical drawings, and any PDF or markdown source.

Question 7

How fast is Nia?

Accepted Answer

Universal search returns sub-5s responses across millions of indexed pages and GitHub files. agentsearch boots in ~100ms when locally cached, ~2s when docs are already indexed on the backend, and 30–120s for a cold index of a brand-new site. Tiny compression models compress search results in milliseconds for token-efficient agent responses.

Question 8

Do I need to reindex when content changes?

Accepted Answer

No. We compute hashes and continuously monitor your sources. Everything stays up to date automatically.

Question 9

How does Oracle mode differ from regular search?

Accepted Answer

Oracle is a research agent that chains multiple searches and builds a dynamic knowledge tree for complex questions. Regular search returns top results; Oracle keeps reading until it finds a real answer.

Question 10

What's the difference between Tracer and Oracle?

Accepted Answer

Oracle is a deep research agent that runs over your indexed sources, chains multiple searches, and builds a dynamic knowledge tree for complex questions. Tracer is an autonomous GitHub search subagent that runs over public repos without indexing first — give it a question and a list of repos, and it traverses them live. Use Oracle for thorough research over your knowledge base; use Tracer for fast lookups against any public GitHub repo.

Question 11

Can I share indexed sources with my team?

Accepted Answer

Yes. Public indexed sources are shared across all Nia users globally — index once, benefit everyone. Private sources stay scoped to your team or workspace.

Question 12

What agents and IDEs does Nia work with?

Accepted Answer

Any agent that can call an API, MCP server, or skill — including Claude Code, Cursor, Windsurf, Cline, Continue, Zed, Codex, Gemini CLI, and OpenCode. Pick whichever interface fits your runtime: HTTP API, CLI, MCP server, or agent skill.

Question 13

What's the difference between Nia Cloud, the CLI, MCP, and agentsearch?

Accepted Answer

Nia Cloud is the hosted platform where indexes live. The CLI, MCP server, and agent skill are different ways for an agent to talk to it. agentsearch is a zero-install front door: 'npx nia-docs <url>' to browse any docs site as a filesystem without an account.

Question 14

Is there a Python or TypeScript SDK?

Accepted Answer

Yes — first-class SDKs for both Python ('pip install nia-ai-py') and TypeScript ('npm install nia-ai-ts'). Both wrap the REST API with built-in retries, streaming, and type safety. See the SDK Quickstart in the docs for examples.

Question 15

What's the difference between free and paid plans?

Accepted Answer

Free includes 50 queries/month, 20 web searches, 5 saved contexts, and 3 lifetime indexings — enough to try every capability. Paid plans (Builder $15/mo, Team $50/seat, Business $99/seat, Enterprise) raise those limits, unlock Oracle / Tracer / Deep Research quotas, and add team features. Credit packs are available on every plan if you need a top-up.

Question 16

How to contact Nia support?

Accepted Answer

Reach us at founders@nozomio.com or join our Discord community.

Careers

Product

Developers

Company

Legal