Inspiration We work in sales and experience firsthand how much time is wasted on repetitive administrative tasks. Hours disappear into email threads, Slack messages, CRM updates, invoice creation, and follow ups. The lack of intelligent automation slows down revenue generating work. We built NOGE to solve the exact problem we face every day.
What It Does NOGE is an autonomous sales agent that executes operational tasks on your behalf. You send a simple message or voice instruction. The agent interprets the request, determines the required actions, and carries them out across your tools. Whether it is sending an invoice to Bob, following up with Taylor so you do not lose your best lead, or updating records, the agent handles the repetitive work so you can focus on selling. It reduces friction between intent and execution.
How We Built It We connected each external service to a central agent node using Composio MCP. For inference, we integrated Cerebras through the OpenAI SDK using LiteLLM, enabling throughput of up to 3000 tokens per second. This gives us fast, low cost, and scalable model responses. We designed a structured memory workflow so the agent retains relevant context across interactions while maintaining clarity and task accuracy.
Challenges We Faced We encountered several technical and product challenges: Voice models were difficult to deploy on Expo and required experimentation to run reliably on device. Demonstrating how the agent works in a simple and intuitive way required careful UX decisions. Enabling smaller models to complete complex multi step tasks required structured prompting and tool constraints. Building effective memory was non trivial. We had to determine what information to store, how to retrieve it meaningfully, and how to inject it back into context in a way that improved performance rather than adding noise. Maintaining tone and context across human machine interactions required careful orchestration.
Accomplishments We Are Proud Of A working memory workflow that meaningfully improves agent performance. Stripe integration that enables accurate transactional flows. Reliable and fast tool calling across services. Strong system performance with low latency.
What We Learned We learned the importance of real market research and constant user conversations. Solving interdisciplinary problems requires both business insight and engineering depth. By combining sales experience with technical execution, we built a product that addresses a genuine day to day pain point.
What Is Next for NOGE We are launching next week. We already have inbound interest from Fabian’s workplace and multiple industry contacts we personally interviewed during development. Next steps: Launch landing page Implement Stripe paywall Open waitlist Begin targeted distribution We have strong marketing capability and a combined audience of tens of thousands of followers. Sales, we are just getting started.
Log in or sign up for Devpost to join the conversation.