Features
Drag-and-drop visual multi-agent workflow builder
One-click export to Python scripts or Jupyter notebooks
Runtime agent spawning via LLM intent routing over MQTT
Actor-model concurrency with isolated async mailboxes
Live dashboard for agent health, cost meters, and error alerts
Auto-persistence — agent state saves on every tick, survives crashes
Multi-LLM support: OpenAI, Anthropic, Google, Mistral, AWS Bedrock, etc.
Community Hub to share and fork workflows
Edge and IoT ready — supports Docker, systemd, native Python runner
Multi-interface: REST, WebSocket, Discord, WhatsApp, Telegram, CLI
Per-agent and aggregate LLM cost tracking
Built-in agents: MainActor, MonitorAgent, DynamicAgent, PlannerAgent, HomeAssistant agents, InstallerAgent, CatalogAgent
MQTT pub/sub as inter-agent messaging backbone
Streaming responses with rolling history summarization
Deploy remote nodes via SSH using InstallerAgent
Multi-agent orchestration
Low-code visual interface
Human-in-the-loop interactions
Conversation pattern customization
Web-based agent monitoring
Group chat and debate patterns