The private alternative to Perplexity.
Search the web. Get private answers.
A native iOS 26 / macOS 26 answer engine that combines web search with private AI. AISight searches the web via a privacy-respecting SearXNG instance (default provided, or self-host your own) and synthesizes answers using Apple's FoundationModels framework (Apple Intelligence) — by default all AI processing happens privately on your device; if you choose it in Settings (iOS/macOS 27+), answers are generated by Apple Private Cloud Compute instead. Zero external dependencies.
Every major answer engine — Perplexity, Google AI Overviews, Bing Copilot — sends your queries to cloud AI services, building a profile of what you search. AISight takes a different approach: by default the AI runs entirely on your device via Apple Intelligence, so your questions never leave your phone or Mac. Only if you opt in, answers are generated by Apple Private Cloud Compute. You control the search backend too (self-hosted SearXNG), meaning no single company sees both what you search and what you ask about it. It's the answer engine for people who think privacy shouldn't be a premium feature.
- Answer factual, encyclopedic, and "how-to" questions with cited sources
- Deep Search — multiple AI research passes for complex questions
- Search the web via SearXNG (aggregates Google, Bing, Brave, and more)
- On-device AI synthesis by default — your data is not sent to a cloud AI service unless you opt in to Apple Private Cloud Compute
- Stream responses token by token with inline citations
- Reciprocal Rank Fusion (RRF) for multi-engine result ranking
- Persist query history locally via SwiftData
- Universal app — runs natively on iPhone, iPad, and Mac
Bring your own SearXNG server → everything is free. Unlimited searches, Deep Search, all features — no purchase needed. You're running your own infrastructure, so there's nothing to pay for.
AISight Pro ($4.99, one-time) unlocks unlimited searches and Deep Search on our default server — and supports indie development.
- Multi-hop reasoning beyond Deep Search (on-device model is ~3B parameters)
- Real-time news with high freshness guarantees
- Complex math or coding assistance
- Image understanding (text-only)
- Work without internet (search requires connectivity)
- Run on devices older than iPhone 15 Pro / iOS 26
- Xcode 26+
- iOS 26.0 / macOS 26.0 deployment target
- Apple Intelligence enabled on device
- A SearXNG instance (local Docker or remote)
A Docker Compose setup is included for local development:
cd searxng
docker compose up -dThis starts SearXNG on http://localhost:8888 with Google, Bing, Brave, and Wikipedia enabled.
For production, deploy your own instance following the SearXNG Docker guide.
- Open
AISight/AISight.xcodeprojin Xcode 26 - The app ships with a default SearXNG instance — for local development, change the URL to
http://localhost:8888in the app's Settings tab or inAISight/App/AppConfig.swift - Select your target (iPhone simulator or My Mac) and Cmd+R
macOS note: Add the "Outgoing Connections (Client)" capability in Signing & Capabilities for network access.
Physical device note: Replace
localhostwith your Mac's local IP in the app's Settings tab.
User Query → SearXNG (search) → ContentFetcher (HTML→text) → FoundationModels (on-device AI by default) → Streamed Answer with Citations → SwiftData (history)
| Layer | Path | Responsibility |
|---|---|---|
| App | AISight/App/ |
Entry point, config, global state |
| Core/AI | AISight/Core/AI/ |
FoundationModels session, Deep Search pipeline |
| Core/Search | AISight/Core/Search/ |
SearXNG API client, RRF ranking, models |
| Core/Fetching | AISight/Core/Fetching/ |
URL → clean text extraction |
| Core/Persistence | AISight/Core/Persistence/ |
SwiftData models and store |
| Features | AISight/Features/ |
Search, History, Onboarding, Settings |
| UI/Components | AISight/UI/Components/ |
CitationText, SourceCard, etc. |
Tech stack: Swift 6, SwiftUI, FoundationModels, SwiftData, URLSession. No external packages.
AI inference runs on-device via Apple Intelligence by default. Data that leaves your device:
- Search queries sent to a SearXNG instance
- HTTP requests to source URLs for content fetching
- Only if you enable "Apple Private Cloud Compute" in Settings: your query and search results, sent to Apple Private Cloud Compute to generate the answer
- Only if you configure a custom OpenAI-compatible provider (iOS/macOS 27+) and confirm the prompt: your question, the text extracted from web sources and the AI instructions, sent to that provider under its own terms (Deep Search makes several requests). The API key is stored in the device Keychain, bound to that provider's host, and removed by Delete All Data.
No analytics. No tracking. No third-party AI services unless you opt in to a custom provider (Apple Private Cloud Compute is Apple's own service and is opt-in).
Contributions welcome. Open an issue before submitting large PRs.
Built by a solo developer in Spain. Entire infrastructure costs under €5/month.


