Rosie

16,670 stars for infra that lets AI agents actually use computers. Click buttons. Type text. Control macOS, Linux, Windows. One API for any OS — cloud or local. Sandboxes, SDKs, benchmarks. Fully open source. #AIAgents #ComputerVision #DevTools 🔗 Link in the comments

Shipping voice features means sending audio to the cloud. Supertonic runs TTS entirely on-device. 31 languages. Sub-100ms latency. Zero API calls. Your data stays yours. #VoiceAI #LocalAI #SelfHosted 🔗 Link in the comments

Your AI setup is scattered across 5 different tools. Personal AI Infrastructure unifies everything into one agentic system. Version-controlled. Self-hosted. Built in TypeScript. #AIAgents #SelfHosted #DevTools 🔗 Link in the comments

Turn any AI agent into a research scientist. Cancer genomics. Drug discovery. Medical imaging. Molecular dynamics. Time series forecasting. 135 skills across 10+ scientific domains. #AIAgents #RAG #DevTools 🔗 Link in the comments

Your AI agent built the wrong thing again. /grill-me forces it to ask detailed questions about what you actually want — before it starts. 77k devs use these Claude skills daily. Small, composable, hackable. #AIAgents #CodeAssistant #Claude 🔗 Link in the comments

Most linters catch syntax. React Doctor catches bad React. Scans for dead code, unsafe patterns, and performance traps your agent missed. Posts findings to PRs. Runs offline. #CodeAssistant #AIAgents #DevTools 🔗 Link in the comments

SoundSource costs $39/year for per-app volume control. FineTune does it all — 10-band EQ, multi-device routing, headphone correction. Free. Open source. One click to install. #macOS #MenuBarApp #OpenSource 🔗 Link in the comments

118+ integrations. One-click OAuth. Auto-syncs every 20 minutes. Runs entirely on your machine. No config. No prompting loops. No cloud. #PrivateAI #LocalAI #AITools 🔗 Link in the comments

Your Downloads folder is a disaster. Files everywhere. No system. Binky watches it, sorts automatically, and holds sketchy stuff for review first. Native. One tap to clean up. #macOS #FileManager #Automation 🔗 Link in the comments

Running local LLMs on your Mac keeps stopping because KV cache fills up too fast. oMLX fixes this: hot models in RAM, cold ones on SSD — swap seamlessly mid-conversation without losing context. Menu bar controls everything. #LocalAI #LLM #SelfHosted 🔗 Link in the comments

Before: AI agents guessing what's in your browser. Now: they see network requests, console errors, and performance traces. MCP server that gives coding agents full Chrome DevTools access. #AIAgents #DevTools #LLM 🔗 Link in the comments