ADC is a desktop application for orchestrating teams of autonomous coding agents across multiple projects. It provides a unified command center to spawn, monitor, and coordinate AI agent teams — with automated QA review loops, workflow-driven task boards, integrated terminals, git workflows, and productivity tools.
User Story : "As a developer managing multiple codebases, I want to delegate tasks to AI agent teams and track their progress visually, so I can ship features faster while maintaining oversight and quality through automated QA."
Architecture Overview
Layer
Components
Technology
Renderer
React Components, TanStack Router, React Query, Zustand
React 19, TypeScript
Preload Bridge
Typed IPC Context, window.api
Electron contextBridge
Main Process
IPC Router, 29 Services, OAuth, MCP Servers, PTY
Node.js, Electron 39
External
Hub Server, GitHub/Spotify/Calendar APIs, Anthropic SDK
REST, WebSocket, OAuth2
Storage
SQLite (Hub), JSON Files (Local), Task Specs
File system, SQLite
Data Flow : React Query hooks call ipc() → Preload bridge → IPC Router → Services → External APIs/Storage
Agent Team Orchestration & Control
Feature
Description
Agent Team Management
Spawn, pause, resume, and terminate multiple Claude CLI agents simultaneously
Workflow-Driven Task Table
Sortable, filterable task table with customizable agent workflows and /implement-feature skill integration
Agent Queue
Queue tasks for sequential agent execution with dependency management
Progress Watching
Real-time sync of docs/progress/*.md files to Hub for crash-safe tracking
Task Launcher
Launch Claude CLI sessions directly from task rows with project context
Automated QA Loops & Testing
Feature
Description
QA Review Pipeline
Automated quality gates — every task reviewed by QA agents before merge
Codebase Guardian
Structural integrity checks for architecture compliance, import health, and type safety
Test Gate Enforcement
Mandatory test suite runs before any work is claimed complete
QA Agents
AI-powered code review, regression detection, and standards enforcement
Feature
Description
Multi-Project Support
Manage multiple codebases with instant project switching
Workspaces
Group related projects with shared settings, max concurrency limits, and device assignment
Device Sync
Multi-device support via Hub — workspaces can be hosted on specific devices
Git Worktrees
Parallel development with visual worktree management
Branch Merging
Visual conflict resolution and merge preview
Productivity & Integrations
Feature
Description
Integrated Terminals
Multi-pane terminal grid (xterm.js + node-pty)
Daily Planner
Time blocking with drag-and-drop scheduling
Daily Briefing
AI-generated summaries and task suggestions
Notes & Ideas
Quick capture with tags and pinning
Spotify Controls
Music playback without leaving the app
Google Calendar
View and create calendar events
GitHub Integration
PRs, issues, and repo management
Slack/Discord
MCP-powered communication tools
Feature
Description
Persistent Assistant
Built-in Claude assistant with conversation history (Anthropic SDK)
Smart Task Creation
Natural language task decomposition
Chrono Time Parser
Parse "tomorrow at 3pm" into timestamps
Voice Interface
Speech-to-text input and text-to-speech output
Screen Capture
Quick screenshots for context sharing
Email Integration
SMTP-based notifications
Notification Watchers
Background monitoring for Slack/GitHub updates
Workflow-Driven Task Table
The Task Table provides sortable, filterable task management with customizable agent workflows:
graph LR
subgraph TaskTable["Task Table"]
Backlog["Backlog"] --> Queue["Queue"]
Queue --> InProgress["In Progress"]
InProgress --> AIReview["AI Review"]
AIReview --> HumanReview["Human Review"]
HumanReview --> Done["Done"]
end
subgraph AgentWorkflow["/implement-feature Workflow"]
Plan["Phase 1: Plan"] --> Spawn["Phase 2: Spawn Agents"]
Spawn --> Execute["Phase 3: Execute in Waves"]
Execute --> Test["Phase 4: Test Gate"]
Test --> QA["Phase 5: QA Verification"]
QA --> Merge["Phase 6: Integration"]
end
TaskTable -.-> AgentWorkflow
Loading
How It Works :
Create a task in the Task Table with requirements and priority
Launch /implement-feature skill from the task row actions
Agents are spawned in waves (Schema → Service → IPC → Components)
Mandatory test suite runs before any work is claimed complete
Progress syncs to docs/progress/*.md for crash recovery
QA agents verify each component before integration
Layer
Technology
Desktop
Electron 39
UI
React 19, TanStack Router 1.95
State
React Query 5, Zustand 5
Styling
Tailwind CSS 4, Radix UI
Validation
Zod 4
Terminal
xterm.js 6, @lydell/node-pty
Backend
Fastify 5, SQLite (Hub)
Testing
Vitest, Playwright
git clone https://github.com/ParkerM2/Agentic-Desktop-Command.git
cd Agentic-Desktop-Command
npm install
npm run dev
Command
Description
npm run dev
Start development mode
npm run build
Production build
npm run lint
ESLint check
npm run typecheck
TypeScript check
npm run test
Run test suite
src/
├── main/ # Electron main process (29 services)
├── preload/ # IPC context bridge
├── renderer/ # React app (25 features)
└── shared/ # Types + IPC contract (single source of truth)
hub/ # Optional sync server (Fastify + SQLite)
.claude/agents/ # 27 specialist agent definitions
Multi-device sync and real-time collaboration:
graph LR
Desktop1["Desktop 1"] <--> Hub["Hub Server"]
Desktop2["Desktop 2"] <--> Hub
Hub --> DB["SQLite"]
Hub --> WS["WebSocket"]
Loading
cd hub && npm install && npm run dev
# → http://localhost:3200
Metric
Value
TypeScript files
~300
Feature modules
25
Main services
29
IPC handlers
30
Agent definitions
27
Color themes
7 × 2 modes
AGPL-3.0 — see LICENSE
Built by Parker · Powered by Claude