IRAB (IRIS Agentic Browser) is a next-generation, local-first browser built for researchers, developers, and power users who want more than just tabs and search.
It transforms the traditional browsing experience into an intelligent, action-driven workflow — where searching, reading, extracting, and automating happen seamlessly inside one system.
Not a browser — a thinking system.
IRAB is part of the IRIS Project Series, a growing ecosystem of advanced AI systems designed to move beyond passive interfaces.
- From tools → to intelligent systems
- From input/output → to autonomous workflows
- From apps → to operating layers
- IRIS → Core AI Assistant
- IRAB → Agentic Browser (this project)
- (Future modules expand the ecosystem)
Traditional browsers:
- You search 🔍
- You open tabs 🌐
- You manually read and extract 📄
IRAB:
- You describe a goal 💬
- It searches, opens, analyzes, and extracts 🤖
- You get structured results instantly 📊
IRAB is built as a multi-layer system:
- Built on Chromium (via Electron)
- Fully custom UI (IRIS-style system interface)
- Supports:
- Tabs
- Address bar
- Profiles
- Web sessions (Google, Gmail, etc.)
- Powered by SearXNG (self-hosted)
- Aggregates results from multiple engines
- No API limits, no tracking, fully local-first
Flow:
User Query → SearXNG → Results → IRAB Processing
IRAB uses a hybrid AI architecture:
- Gemma 4 (via Ollama)
- Handles:
- Summarization
- Extraction
- Basic reasoning
- Offline workflows
- Groq (LLaMA 3.1 / 3.2)
- Handles:
- Fast responses
- Ranking
- Lightweight reasoning
- Gemini API
- Handles:
- Screenshot understanding
- Complex UI interactions
IRAB converts simple queries into multi-step workflows:
Search → Open → Extract → Analyze → Summarize → Export
Example:
“Compare top AI agent tools in 2026”
IRAB will:
- Search across sources
- Open multiple pages
- Extract key data
- Build a comparison
- Return structured output
-
Extracts:
- Text
- Tables
- Structured data (JSON)
-
Works across multiple tabs
-
Supports multi-page aggregation
-
Multiple browser profiles (like Chrome)
-
Persistent sessions:
- Gmail
- Any logged-in platform
-
Isolated environments:
- Research
- Work
- Personal
- Form filling
- Repetitive task execution
- Macro-based workflows (planned)
- 🟢 Local-first AI (zero cost core usage)
- ⚡ Smart search → structured answers
- 📊 Multi-page data extraction
- 🤖 Agent-style workflows
- 🌐 Real browser environment (Chromium)
- 👤 Persistent profiles
- 📁 Export to Markdown / CSV
- 🧩 Modular system architecture
IRAB follows the IRIS System Design Language:
- Dark system UI
- Neon green accent (growth + control)
- Glass panels + subtle glow
- OS-style interaction model
- Minimal, focused, powerful
- Frontend: Electron + Vite + React
- Browser Engine: Chromium (via Electron)
- Automation (future/optional): Playwright
- Search: SearXNG (self-hosted)
- Local AI: Ollama + Gemma 4
- Cloud AI: Groq + Gemini
IRAB is designed with realism:
- Not trying to replace Chrome
- Not trying to replace Google
Instead:
It sits on top of the web and makes it intelligent
- Multi-agent orchestration
- Deep research mode (local RAG)
- Visual automation (click-by-vision)
- Macro recording & replay
- IRIS system integration (cross-app workflows)
IRAB is not just about browsing faster.
It’s about:
- Thinking faster
- Extracting faster
- Acting faster
“Search is the input. Intelligence is the interface.”