-
-
The UI of YouTube GH CreatorOS agent
Inspiration
Growing as a creator is hard. Not because people lack talent or ideas, but because they lack direction. A proper growth strategy is something only creators with big budgets or the right connections can access. We wanted to change that. YouTube GH CreatorOS started from a simple question: what if every creator had access to a personal strategist that never sleeps, never charges a retainer, and gets smarter every time you talk to it?
What it does
YouTube GH CreatorOS is an autonomous AI agent that turns a creator's goals into a personalized, research-backed growth roadmap and delivers it straight to their inbox.
The creator shares their niche and goals through a chat interface. From there the system takes over. A Researcher Agent digs into current trends and opportunities in their space. A Creation Strategist Agent takes that research, combines it with the creator's history and context, and builds a roadmap tailored specifically to them. A human reviews and approves the roadmap before it goes anywhere. Then it lands in the creator's inbox, ready to use.
The more a creator interacts with it, the better it gets. Persistent memory means it remembers who you are and what you are working toward, so you never have to repeat yourself.
How we built it
We built the entire pipeline on Airia, designing two agents with distinct roles. The Researcher Agent focuses purely on market intelligence and trend analysis. The Creation Strategist handles synthesis and roadmap generation, drawing on both the research output and the creator's stored memory. A custom Python node handles email delivery through Gmail, with logic to handle different SMTP configurations cleanly. And a human approval step sits in the middle of it all to make sure nothing goes out that is not ready.
Challenges we ran into
Getting context to pass cleanly between nodes was harder than expected. The creator's email address had to travel from the chat input through LLM-generated content all the way to the Python delivery node without getting lost. We also spent real time debugging SMTP, specifically handling the difference between SSL and STARTTLS connections so the email node would not silently fail. And we had to be intentional about pipeline order, making sure the approval step came before delivery and not after.
Accomplishments that we are proud of
We shipped a fully working end-to-end pipeline. A creator can come in, share their goals, and get a personalized roadmap delivered to their inbox with a real human reviewing it in between. The memory layer actually works and makes the experience feel personal. And the two-agent architecture holds up, each agent stays in its lane and the output is better for it.
What we learned
Orchestration is everything in multi-agent systems. You can have great individual agents and still have a broken product if the pipeline connecting them is not solid. We also learned to stop adding features when time is short and focus on making what exists work really well. The human-in-the-loop step taught us something too. It is not just a safety net. It genuinely improves the quality of what gets delivered.
What is next for YouTube GH CreatorOS
We want to add Google Sheets integration so every creator interaction is logged and trackable over time. PDF generation is on the list so roadmaps arrive as polished documents rather than plain text. We are also thinking about niche-specific agents built for YouTube, Instagram and TikTok separately, and eventually a dashboard where creators can track which parts of their roadmap are actually moving the needle.
Log in or sign up for Devpost to join the conversation.