Show HN: Enclyra – 3-4B local models that do real work on an 8 GB Mac, offline

Category: ai-ml

Tags: local-ai, macos, offline, agent, privacy

Score: 6.8/10 (Innovation: 7, Technical: 7, Documentation: 6, Utility: 7)

Enclyra is a macOS desktop AI assistant that runs small local models (3–4B) offline to perform multi-step file, media, and system tasks via natural language commands. It combines a local LLM planner with 80+ core actions, undo support, and on-device memory, targeting privacy-conscious users on 8 GB Apple Silicon Macs. The project is notable for its honest reliability reporting and clear privacy disclosures, though it remains a closed-source beta.

Target audience: macOS power users, privacy-conscious consumers, and developers interested in local AI agents

Repository: https://www.enclyralabs.com

View on Hacker News