Native macOS AI workspace for deep context analysis, dynamic DAG automations, GUI actions, and Model Context Protocol (MCP) tool integration.
View Repository on GitHub
Bridging desktop workflows with Google Gemini 3.7 models, live voice streams, and MCP servers.
Dynamic model toggling between 3.7 Flash for automation speed and 3.1 Pro for deep reasoning. Generates text, code, images, videos (Veo 3.1), and audio (Lyria 3).
Build Directed Acyclic Graphs visually. Chain Input, AI Prompt, API Request, Python Code, MCP Tool, and Router nodes seamlessly.
Cloud vector store indexing (file_search_stores) with multi-query semantic expansion and chronological "Connect Dots" map-reduce re-ranking.
Connect external stdio or REST/SSE MCP servers (Pythia Oracle, Financial Datasets, DeepWiki). Automates conversion into native Gemini function declarations.
Native WebSocket integration delivering real-time speech-to-text dictation and duplex interactive voice communication with low latency.
Integrated ScreenCaptureKit window capture for instant visual audits alongside directory drop-folder watchers that trigger visual DAG flows.
Fine-tune reasoning turns with Min, Low, Med, and High thinking budgets directly per execution thread.
Configure vector search recall from 1 to 50 chunks with multi-query variations enabled on demand.
Live telemetry displaying active request counts, total tokens consumed, time saved, and heaviest query profiles.
Built with SwiftUI, SwiftData, AppKit, and ScreenCaptureKit paired with an async Python 3 backend engine.