How codejunkie99/agentic-stack-desktop works
Agentic Stack is a local-first macOS workspace and backend that unifies coding-agent conversations, project work, and reusable memory. The main workflow runs from a user composing work in the desktop app through its bridge to workspace services, which coordinate model selection and agent execution; approved knowledge and conversation evidence can then be searched or recalled across tools. The README documents additional loops, terminals, setup, and operations. Several backend paths are represented from documented capabilities and the tree; unsampled internal wiring is left unspecified.
Main entry points: AppModel.swift App Model · service.py Workspace Service · context_mcp.py Context MCP · Bridge.swift Backend Bridge
Components 22
macOS Workspace
- Desktop App
apps/macos/Sources/AgenticWorkspaces/AgenticWorkspacesApp.swift - App Model
apps/macos/Sources/AgenticWorkspaces/AppModel.swift - Conversation UI
apps/macos/Sources/AgenticWorkspaces/ConversationView.swift - Knowledge Views
apps/macos/Sources/AgenticWorkspaces/KnowledgeGraphView.swift - Backend Bridge
apps/macos/Sources/AgenticWorkspaces/Bridge.swift
Workspace Backend
- HTTP Bridge
harness_manager/workspaces/server.py - Workspace Service
harness_manager/workspaces/service.py - Model Router
harness_manager/workspaces/model_router.py - Workspace Store
harness_manager/workspaces/store.py - Task Streams
harness_manager/workspaces/task_stream.py - Loop Coordinator
harness_manager/loops/runner.py - Loop Checkpoints
harness_manager/loops/storage.py
Shared Context
- Knowledge Import
harness_manager/workspaces/knowledge.py - Personal Memory
harness_manager/workspaces/personal_memory.py - Context MCP
harness_manager/context_mcp.py
Projects and Operations
- Project Setup
harness_manager/workspaces/setup.py - Interactive Terminals
harness_manager/workspaces/terminal.py - Mission Control
harness_manager/mission_control.py
Other components
- User
- Agent Histories
- Coding Agents
- Private Server
Connections 23
- User → Desktop App: uses
- Desktop App → App Model: hosts
- App Model → Conversation UI: presents
- Conversation UI → App Model: submits work
- App Model → Backend Bridge: sends requests
- Backend Bridge → HTTP Bridge: RPC requests
- Private Server → HTTP Bridge: hosts service
- HTTP Bridge → Workspace Service: dispatches
- Workspace Service → Workspace Store: reads and writes
- Workspace Service → Model Router: selects model
- Workspace Service → Task Streams: streams progress
- Workspace Service → Coding Agents: runs tasks
- Workspace Service → Knowledge Import: imports approved sources
- Workspace Service → Personal Memory: manages profile
- Knowledge Views → App Model: requests graph data
- App Model → Backend Bridge: requests knowledge
- Context MCP → Agent Histories: reads history
- Context MCP → Workspace Store: reads indexed memory
- Coding Agents → Context MCP: recalls context
- Loop Coordinator → Loop Checkpoints: stores checkpoints
- App Model → Project Setup: requests setup
- App Model → Interactive Terminals: controls terminal
- App Model → Mission Control: requests operations