Hi HN, I've been building Slate for the past few months and just open-sourced it. It's a native macOS app that puts AI chat and web browsing in the same window. The idea came from how I actually use AI day to day. I'd ask Claude or GPT something, get a bunch of links or recommendations, then cmd-tab to a browser, open tabs, lose context, and go back and forth. It felt broken. I wanted the browser inside the AI conversation, not the other way around. So Slate is an AI workspace first, browser second. You ask a question, get a synthesized answer, click into sources right there in the same session, and keep asking follow-ups with full context. Each tab has its own conversation history. A few things that might be interesting to this crowd: - Native SwiftUI + WebKit, no Electron. Runs light on memory and battery on Mac. Downside: it's MacOS only, but it's the choice i took to create best possible UI and the user experience. - Most other "AI Browsers" bolt AI onto the chromium as an after-thought, but I wanted to build AI-first experience. - Supports Anthropic, OpenAI, Gemini, and Ollama for local models. Bring your own API keys. - Built-in content blocking, so pages load fast and clean. - Sessions group your tabs into workspaces you can archive and restore. - MIT licensed. It's very much a v1. There's plenty I want to improve — extensions, better history, sync, AI Agents, MCPs, etc. — but the core loop of chat → browse → follow up feels solid enough to share. Repo: https://github.com/slate-ai/slate Download (.dmg): https://github.com/slate-ai/slate/releases/download/v1.0.0/s... Would love feedback, especially from anyone who's worked with WebKit or has opinions on how AI should integrate with browsing. Happy to answer questions about the architecture or anything else. |