How maxmasai/agentic-web works
Agentic Web is a PySide6 desktop platform for multi-agent AI orchestration and autonomous desktop/browser operations. Its primary flow is a user task entering the desktop application, reaching the orchestration runtime, and being delegated to specialist agents and tools before producing a result. The repository also documents canvas workflows, MCP, skills, plugins, token analytics, and live voice. The graph combines the supplied excerpts, SOURCE INDEX references, and README claims; several UI and feature implementations were not sampled, so their internal wiring is not asserted.
Main entry points: main.py Task orchestrator · llm_provider.py Model provider · app.py Desktop shell · browser_controller.py Browser automation
Components 23
Desktop application
- Desktop shell
app.py - Mission interfaces
gui/pages - Token dashboard
gui/pages/token_manager_page.py
Agent orchestration
- Task orchestrator
core/system/main.py - Agent roster
core/agents/agentlist.py - Agency catalog
core/agency_agents_manager.py - Skills manager
core/skills_manager.py - Model provider
services/llm_provider.py
Execution capabilities
- Browser automation
browser/browser_controller.py - Autonomous canvas
core/workflows/canvas_engine.py - Internal tools
core/system/internal_tool_executor.py - Live voice
voice/backend/server.py - Plugin manager
plugins/plugin_manager.py - MCP tools
services/mcp_client.py
Platform services
- Authentication
services/auth_service.py - Token tracking
services/token_manager.py - Persistent state
system/state_store.py - Window layouts
system/fancyzones_manager.py
External systems
- Model APIs
- Websites
- Desktop OS
- MCP servers
Other components
- Desktop user
Connections 25
- Desktop user → Desktop shell: starts task
README.md - Desktop shell → Task orchestrator: starts runtime
app.py - Desktop shell → Mission interfaces: shows pages
app.py - Mission interfaces → Task orchestrator: submits tasks
- Task orchestrator → Agent roster: selects agents
core/system/main.py - Task orchestrator → Browser automation: runs browser tasks
core/system/main.py - Task orchestrator → Internal tools: executes tools
core/system/main.py - Task orchestrator → Window layouts: tiles windows
core/system/main.py - Task orchestrator → Model provider: requests completion
- Task orchestrator → Agency catalog: uses specialists
README.md - Agency catalog → Skills manager: integrates skills
core/agency_agents_manager.py - Task orchestrator → Autonomous canvas: supports workflows
README.md - Task orchestrator → Live voice: offers voice mode
README.md - Task orchestrator → Plugin manager: supports plugins
README.md - Task orchestrator → MCP tools: dispatches MCP tools
README.md - Model provider → Authentication: checks credentials
services/llm_provider.py - Authentication → Model APIs: authenticates
services/auth_service.py - Model provider → Model APIs: sends prompts
services/llm_provider.py - Browser automation → Websites: controls pages
browser/browser_controller.py - Browser automation → Desktop OS: opens applications
README.md - Window layouts → Desktop OS: arranges windows
README.md - MCP tools → MCP servers: connects tools
README.md - Token dashboard → Token tracking: reads usage
gui/pages/token_manager_page.py - Token tracking → Model provider: tracks consumption
- Task orchestrator → Persistent state: persists agent state
README.md