Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/jasonkneen/openclicky/llms.txt

Use this file to discover all available pages before exploring further.

OpenClicky is a native macOS menu-bar app built by Jason Kneen that puts Clicky — an AI companion powered by Claude and other LLMs — permanently at your fingertips. Unlike chat-based AI tools, OpenClicky lives in your menu bar, responds to voice, sees your screen on demand, and can autonomously complete multi-step tasks through its Agent Mode.

Why OpenClicky

OpenClicky combines voice, vision, and agentic automation in a single native macOS app:
  • Voice-first: Push-to-talk or wake word activation — no typing required for most interactions
  • Screen-aware: Clicky can see what you’re looking at and point to specific UI elements with a native cursor overlay
  • Agent-capable: For complex tasks, Clicky spawns a Codex-powered agent that can run shell commands, use bundled skills, and work autonomously
  • Fully local: Your API keys stay on your machine — no cloud sync, no hosted auth, no Google login
  • Extensible: A rich skill system with bundled skills and support for custom skills

Architecture Overview

OpenClicky is a pure macOS app using SwiftUI and AppKit:
  • App type: Menu-bar companion using LSUIElement=true (no Dock icon)
  • Frameworks: SwiftUI, AppKit, AVFoundation, ScreenCaptureKit
  • Voice input: Push-to-talk via a global CGEvent tap with pluggable transcription providers
  • AI responses: Claude via the Anthropic API with your own key
  • Text-to-speech: ElevenLabs, Cartesia, Deepgram, OpenAI Realtime, or Microsoft Edge via local key configuration
  • Screen context: ScreenCaptureKit screenshots when you invoke help
  • Agent Mode: Bundled Codex runtime with resource pack in AppResources/OpenClicky/

Key Components

CompanionManager

Central app state machine for voice, screen capture, Claude, TTS, overlay, settings, and Agent Mode.

MenuBarPanelManager

Manages the menu-bar icon and floating panel lifecycle.

OverlayWindow

Cursor overlay, agent dock icons, captions, and response cards.

CodexHUDWindowManager

The Agent Mode dashboard window.

What Clicky Can Do

Clicky handles a broad range of tasks from a single conversational interface:
  • Fresh web search for facts and news
  • Screen-aware guidance using [POINT:x,y:label] and [TYPE:x,y:label] directives and native cursor overlay
  • Image gallery display for visual results
  • Child workers and agent spawning for larger, multi-step tasks
  • GitHub integration through Composio MCP
  • Local shell and file work inside configured project roots
  • Frontend builds and previews
  • Reports, PDFs, DOCX files, and spreadsheets
  • Repo scaffolding and day-to-day dev work
  • Native computer-use fallback for GUI automation

Repository Layout

PathContents
cursor-buddy.xcodeprojXcode project file
cursor-buddy/macOS app target source
cursor-buddyTests/Focused app unit tests
cursor-buddyUITests/UI test scaffolding
AppResources/OpenClicky/Bundled model instructions, skills, wiki seed, Codex runtime, and completion audio
Packages/Swift package modules (Browser, Core, Markdown, Memory, UI)
scripts/Release, version bump, appcast generation scripts
docs/Internal documentation
The legacy cursor-buddy folder and scheme names are kept for project continuity. The product, bundle identifier (com.jkneen.openclicky), and display name are OpenClicky.

Next Steps

Installation

Build and run OpenClicky from source.

Configuration

Set up API keys and provider preferences.

Build docs developers (and LLMs) love