Pine Chat is a macOS AI agent app that connects to a local model like LM Studio. Hand it a design document (.md) and hands-off mode will build the app for you, right up to the final build.
Features
More than a chat window: coding, research, and outside integrations run through a single local AI agent.
Connects automatically to LM Studio (port 1234) or any OpenAI-compatible API. A 20B-class model or larger is recommended.
Privacy-respecting web search through SearXNG, which can also run as a background service on macOS.
Give it a design document (.md) and "hands-off" mode will implement the app all the way to the final build.
Load .md files as a reference dictionary (RAG) so the agent can draw on project-specific knowledge.
Connects to Discord, Telegram, and your calendar to summarize activity and hold conversations for you.
Released under GPL-3.0-or-later. Source is freely available to read and modify on GitHub.
How it works
Load a model and start a server in LM Studio (default port 1234).
Download the latest .dmg from the GitHub Releases page and launch it.
It connects to your local AI automatically on launch. Create a project and start building, researching, or just chatting.
npm install then npm start, or run ./build.sh to build a .dmg.