I built Atlarix because I wanted the agentic power of tools like Lovable/Windsurf but without the cloud lock-in or privacy trade-offs. I wanted an AI that actually lives on my machine, understands my local dev environment, and lets me control the models I use.
What makes it different:
Native & Local-First: It’s a standalone desktop app (Electron + Rust/Node), not a VS Code extension. It has direct control over your terminal, file system, and browser to run agentic loops (scaffold -> run -> debug -> fix).
BYOK (Bring Your Own Keys): Complete sovereignty. Use OpenRouter or run local models via Ollama/LM Studio completely offline.
"Lovable-style" Live Previews: It spins up dev servers (Vite, Next.js, Python, Go) and gives you a live, interactive preview and a diagram that the AI can see and debug in real-time.
Agentic Workflow: It doesn't just write code; it runs commands, reads build errors from the terminal, and auto-fixes issues without you copy-pasting stack traces.
Tech Stack:
Frontend: React, Tailwind, Framer Motion
Backend: Electron (Main Process), SQLite (Local Vector DB for RAG)
Agent: TypeScript-based agentic loop with tool use (57+ tools for file ops, searching, terminal control)
It’s currently v1.0. Would love to hear your feedback on the agentic workflow vs. standard chat assistants!
Hi HN, I’m Amariah, the creator of Atlarix.
I built Atlarix because I wanted the agentic power of tools like Lovable/Windsurf but without the cloud lock-in or privacy trade-offs. I wanted an AI that actually lives on my machine, understands my local dev environment, and lets me control the models I use.
What makes it different:
Native & Local-First: It’s a standalone desktop app (Electron + Rust/Node), not a VS Code extension. It has direct control over your terminal, file system, and browser to run agentic loops (scaffold -> run -> debug -> fix). BYOK (Bring Your Own Keys): Complete sovereignty. Use OpenRouter or run local models via Ollama/LM Studio completely offline. "Lovable-style" Live Previews: It spins up dev servers (Vite, Next.js, Python, Go) and gives you a live, interactive preview and a diagram that the AI can see and debug in real-time. Agentic Workflow: It doesn't just write code; it runs commands, reads build errors from the terminal, and auto-fixes issues without you copy-pasting stack traces. Tech Stack:
Frontend: React, Tailwind, Framer Motion Backend: Electron (Main Process), SQLite (Local Vector DB for RAG) Agent: TypeScript-based agentic loop with tool use (57+ tools for file ops, searching, terminal control) It’s currently v1.0. Would love to hear your feedback on the agentic workflow vs. standard chat assistants!