The system is built around multiple specialized sub-agents, each with a clear role — like a small AI production team:

The Producer
the main orchestrator agent. It receives tasks, breaks them down, delegates to the right sub-agents, reviews outputs,
and delivers the final result. It decides which model is best for each job and manages the entire workflow.

The Copywriter 
a dedicated text agent that generates marketing copy, headlines, social media posts, and brand messaging.
It follows tone guidelines and adapts to different formats.

The Visual Creator
an agent trained to generate images and video content. It has specific skills for different visual styles
and formats, producing ready-to-use marketing materials.

The Researcher
handles market research, competitor monitoring, and trend analysis. It gathers data, evaluates findings,
and feeds insights back to the Producer for strategic planning.

All of this runs autonomously through a Telegram interface. I send a high-level request — „prepare a marketing campaign for X“ —
and the system plans the content, writes the copy, generates visuals, and delivers a complete package. No micromanaging.

I built and deployed the entire system solo — server setup, API integrations, agent architecture, prompt engineering,
model selection, monitoring, and maintenance.