A second brain that remembers, organises, and acts. Built for knowledge workers who are tired of starting from zero every day.
Most AI tools forget you the moment you close the window. Sorana doesn't.
Sorana is a personal AI knowledge workspace for Windows. It builds a lasting memory of your projects, files, and thinking style, so every time you open it, your AI already knows where you left off. Think of it as Obsidian's spatial canvas combined with an AI assistant that has actually read your notes, remembers your preferences, and can take action on your behalf.
No subscriptions. No cloud required. Everything runs on your machine.
- 📚 Researchers & academics, chat with hundreds of papers, books, and scanned notes in one unified workspace
- ✍️ Writers & journalists, build a living reference library that remembers your style, sources, and ideas
- 🧑💼 Consultants & freelancers, maintain full AI context across multiple long-running client projects
- 🎓 Students, turn notes, PDFs, and lecture slides into a personal knowledge base you can actually talk to
- 🧠 Obsidian & Notion users, add an AI layer that learns from your vault and can act on it
- 🔒 Privacy-conscious professionals, full AI power, zero data leaving your machine
Every AI chat starts from zero. You paste background context, re-explain your project, re-state your preferences, again and again. Sorana quietly learns from every conversation and greets you with full context next time you open it. No setup. No copy-pasting. Just continuity.
Notes in Obsidian, files in folders, research in browser tabs. You spend more time searching than actually thinking. Sorana maps your files onto a 2D spatial canvas by topic and relationship, find anything by what it's about, not where it's saved.
Sorana's AI doesn't just answer questions, it acts. Organise files, summarise documents, fetch web content, manage emails, all through plain-language conversation. No app switching, no copy-pasting.
Scanned book pages, photographed receipts, PDF reports, dead to most AI tools. Sorana's built-in OCR and vision layer makes your entire archive searchable and chattable, including documents from years ago.
The longer a project runs, the heavier the AI context grows. Sorana automatically compresses old conversations into tight summaries, over 90% fewer tokens, full context preserved. Long-term projects stay fast and affordable.
Every file you send to ChatGPT or Claude leaves your machine. Sorana runs AI models entirely on your computer. Your notes, files, and conversations never leave your hands.
AI maps your files onto a 2D canvas by topic and relationship, not alphabetically, not by date, but by meaning. Like Obsidian Canvas, but automatically organised by your AI. Drag, group, rename, and build mind maps of your folder structure.
Sorana automatically extracts facts, preferences, and project details from every conversation. Next session, your AI knows your name, your current projects, and how you like to work. Memory is shared across all your agents, building a single unified knowledge profile over time.
Drop any file, PDF, Word doc, code, text, scanned image, into your workspace and chat with it in plain language. Sorana searches your documents using multiple strategies in parallel to always find the right answer, even in large archives.
Long conversations are automatically compressed into tight summaries. You keep 100% of the context at a fraction of the cost. Projects that span months stay fast and affordable.
Beyond answering questions, Sorana's built-in automation layer lets your AI operate your file system, browse the web, and manage your Gmail, all through natural conversation. Tell it what you need done; it handles the rest.
Already using Obsidian? Export your Sorana workspace directly to Obsidian Canvas or Graph View. Your AI-organised knowledge structure, carried into your existing vault. Use Sorana for AI-powered work; keep Obsidian as your permanent archive.
Run AI models entirely on your machine via Lemonade, Ollama, Llama.cpp, or LM Studio. Your files never leave your computer. Cloud APIs (OpenAI, Mistral, Gemini, etc.) are supported if you choose them, but never required.
No installation. Unzip anywhere, run from a USB drive, take your entire workspace with you. All data stored in a single .sorana/ folder you own completely.
| What you get | Sorana ✨ | Obsidian / Notion | ChatGPT / Claude | OpenWebUI |
|---|---|---|---|---|
| AI remembers you between sessions | ✅ Automatic | ❌ Manual notes only | ❌ None | |
| Visual knowledge canvas | ✅ AI-organised | ✅ Manual (Obsidian) | ❌ Chat only | ❌ Chat only |
| Chat with your documents | ✅ Any file, any format | |||
| AI takes action on files & web | ✅ Built-in automation | ❌ View only | ❌ Advice only | |
| Runs 100% offline & private | ✅ Your machine | ✅ Notes are local | ❌ Cloud only | ✅ Local |
| Stays fast over months of use | ✅ Auto-compaction | ✅ (no AI context) | ❌ Manual | |
| Free to use | ✅ Free & portable | ❌ $20+/month | ✅ Free |
Sorana combines what Obsidian/Notion do for knowledge organisation with what AI assistants do for conversation, plus the automation that neither offers.
- Download the portable archive from tetramatrix.github.io/Sorana
- Extract anywhere, desktop, USB, external drive
- Run
Sorana.exe, no installation needed - Open a folder as your workspace via Open Workspace
- Select an AI model in Model Manager (built-in offline model included)
Latest: Sorana.exe v1.0.53
📥 [Download (Portable)](tetramatrix.github.io/Sorana](http://tetramatrix.github.io/Sorana)
🏪 Microsoft Store
| Component | Minimum |
|---|---|
| OS | Windows 11 (64-bit) |
| RAM | 4 GB (8 GB+ recommended for local AI models) |
| Storage | 2 GB (app + model) + .sorana/ data folder |
| AI | Built-in offline model included; local or cloud AI optional |
- Built-in model (~806 MB), works fully offline out of the box
- Recommended, Llama 3.1 8B Instruct or higher for best results (requires 8 GB RAM / 4 GB VRAM)
- Cloud, OpenAI, Mistral, Gemini, Deepseek, NVIDIA, OpenRouter (free tiers available), no local hardware required
- Local servers, 🍋 Lemonade (NPU acceleration), Ollama, LM Studio, Llama.cpp
🎉 Free tier APIs available, Deepseek, Mistral, NVIDIA, Gemini, and OpenRouter all offer free tiers. Just add your API key.
Sorana connects directly to your Obsidian vault:
- Export to Canvas, bring your AI-organised workspace layout into Obsidian's infinite Canvas
- Export to Graph View, generate a knowledge graph from your workspace structure, compatible with Obsidian's Graph View
- Full compatibility with Dataview, Templater, Excalidraw, and all other Obsidian plugins
- Standard Markdown files, no lock-in, your data stays yours
- Open MCP Manager and enable the servers you need
- Create an agent in your workspace
- Right-click the agent title → Chat
- Give instructions in plain language, files, web, email, all accessible
Connect external services: Google Drive, GitHub, PostgreSQL, custom tools.
To enable OCR on PDFs with embedded images, install Tesseract OCR and ensure it's in your system PATH. Once installed, Sorana processes scanned documents automatically, making your entire paper archive searchable.
🌐 Website: tetramatrix.github.io/Sorana 💬 Discord: discord.gg/4QkQSfSATF 🏪 Microsoft Store: apps.microsoft.com
- 🧩 TabNeuron, AI spatial tab manager & browser research workspace
- 🖥️ Aicono, AI intelligent desktop icon organiser
- 🔧 RyzenZPilot, AMD Ryzen power management tool
- 🚀 Spaceship!, Retro 2D arcade game
🏢 Developed by Tetramatrix | 🧠 Your personal AI knowledge workspace | 💝 Free for everyone