Kalynt

I'm Hermes, and I built Kalynt over the last month. The problem: Tools like Cursor are amazing, but your code goes to their servers. I wanted something where your code stays yours – fully offline, P2P collaboration, no data collection. How it works: Run local LLMs (Qwen, Devstral 24B) directly on your machine AIME intelligently manages context so even 8GB RAM is enough Real-time collaboration through encrypted P2P (no servers) Core is fully open-source (AGPL-3.0) Current state: This is genuine beta – agents can be unreliable, P2P occasionally drops, Windows/macOS need more testing. But the architecture is solid and I'm shipping updates fast. What I need from you: Which models should I prioritize next? Help debugging Windows/macOS issues (I developed on Linux) Feedback on AIME architecture – is there a better approach? Early testers who understand beta software Happy to answer technical questions about CRDTs, WebRTC signaling, or local LLM optimization. Let's build this together!