The Goddess of Information
Send a link. Drop a document. Record a voice note. Forward an email. Iris organises everything into a searchable knowledge base, automatically, and entirely on your own infrastructure, never sent to a US cloud, never used to train a model, and yours to delete at any time.
Apply for Iris access →Five intake channels, one knowledge base
Everything you send is automatically classified, processed by specialised agents, and archived into a searchable knowledge base.
Links & URLs
Paste any URL. Iris extracts the full article, discovers related links, and checks for duplicates before filing. Works with news, research papers, forums, and social media.
Documents
Upload PDFs, images, or documents. Hush Cyclops extracts the text, OCR included, then Iris classifies and files it into the right category, fully searchable from the moment it lands.
Voice & Audio
Send a voice message. Hush Echo transcribes it on-device, then Iris routes the text automatically, research queries get researched, notes get filed, memos get archived.
Iris monitors your inbox, extracts content from new messages, and archives them directly to your knowledge base. Fully automatic, no forwarding needed.
Text & Notes
Type anything. Iris automatically decides what to do with it: research query, note to file, memory to store, or conversation. No manual sorting needed.
RSS & Feeds
Subscribe to RSS feeds. Iris polls them on schedule, extracts new articles, and archives them automatically. Your knowledge base grows while you sleep.
The Iris pipeline
From raw input to searchable knowledge in seconds. Every step runs locally, nothing leaves your hardware we own.
Intake & Sanitisation
Input arrives via email, web, or messaging. Iris checks it for safety, enforces size limits, and prepares it for classification.
LLM Intent Classification
An AI model classifies each input, research, file, recall, memorise, or chat, with a confidence score. If AI classification is unavailable, keyword matching takes over seamlessly.
Smart Routing
Based on the classification, Iris automatically routes each item down the right path, web research, filing, recall, or safety checks, with no manual sorting.
Processing & Extraction
Each item is processed end to end: web content is fetched and summarised, text is extracted and indexed into folders, and new material is cross-referenced against your existing knowledge.
Archival & Indexing
Everything lands in your knowledge base, fully indexed for text search and semantic search. Every item is tagged, timestamped, and immediately searchable.
Everything built in
No plugins to install. No third-party services to configure. Every feature runs on the same sovereign stack.
Iris vs. cloud knowledge tools
| Iris | Notion AI | Obsidian | Evernote | |
|---|---|---|---|---|
| Runs on your hardware | ✓ | ✗ | ✓ | ✗ |
| LLM-powered auto-classification | ✓ | ✗ | ✗ | ✗ |
| Multi-agent pipeline | ✓ | ✗ | ✗ | ✗ |
| Auto web scraping & archival | ✓ | ✗ | ✗ | ✓ |
| Voice transcription intake | ✓ | ✗ | ✗ | ✓ |
| Email auto-archival | ✓ | ✗ | ✗ | ✓ |
| RSS feed ingestion | ✓ | ✗ | ✗ | ✗ |
| Vector semantic search | ✓ | ✓ | ✗ | ✗ |
| Zero data leaves your network | ✓ | ✗ | ✓ | ✗ |
| GDPR sovereign, no US transfer | ✓ | ✗ | ✗ | ✗ |