I participated in Google's Gemini Live Agent Challenge on Devpost. Everyone was building chatbots. I wanted to show what real multi-agent orchestration could look like — not just one agent answering questions, but dozens of specialists working together. What problem was I trying to solve? Marketers are drowning in SaaS subscriptions. Jasper, Canva, OpusClip, SurferSEO — easily $500+ per month. Worse, your data goes to the cloud, and none of these tools talk to each other. A blog post from one tool doesn't automatically feed into a video script or a social thread. I wanted to prove that a single, self-hosted platform with 70+ specialised agents could replace all of them. And I wanted it to be open source — so marketers own their data and their workflow forever. How did my approach evolve during this launch? Honestly? It started as a messy hackathon prototype. I had 79 agent definitions hardcoded in one file. But as I used it myself — generating brand kits, SEO blogs, Veo 3 video scripts — I realised this needed a proper visual interface. That's when I added the React Flow canvas. Drag, connect, automate. No coding required for campaign workflows. The Asset Library and Analytics dashboard came later — after realising that generated content was useless if you couldn't find it again. This is still a prototype in many ways. But it works. I've already used it to launch test campaigns. And because it's MIT licensed, anyone can fork it, improve it, or build a business on top of it. Would love your feedback. Try the Docker command. Break things. Open issues. Let's build the open-source future of marketing AI together. - Narender

MarketingAI Studio

Open source AI marketing studio with 79 agents

Open source AI marketing studio with 79 agents