Mia launches local AI workspace daemon

Local-first agent tooling only feels real when the remote view is better than SSH, and Mia is leaning into that with a phone-first control plane instead of a cloud relay. The interesting part is whether it stays frictionless when the agent runs for hours and context has to survive restarts.

• –P2P mobile sync is the core differentiator: you get live visibility without routing code or output through a third-party middleman.
• –Support for OpenCode, Claude Code, Gemini CLI, and Codex lowers adoption friction because Mia sits on top of tools people already use.
• –Persistent memory is the sticky feature if it actually preserves useful project context across sessions.
• –Local-model support broadens the appeal for Ollama and llama.cpp users, but setup polish and pairing reliability will decide whether this feels magical or fiddly.