I’ve been building exactly this stack for the past year and a half – n8n workflows orchestrating LLM agents (Claude and OpenAI), with per-client context management, memory layers, and OAuth 2.0 integrations across Google Workspace, CRMs, and third-party APIs.
Most recently, I architected a multi-agent AI system for a digital media agency – state-driven workflows with persistent context profiles per client, structured prompt chains with fallback logic, and automated Google Workspace orchestration. The system handles complex multi-step tasks with explicit state management, not probabilistic guessing.
I’m the founder of Evara AI (incubated at IIT Bhubaneswar), India-based, and I build code-first – custom Python/Node.js backends with n8n as the orchestration layer, not the ceiling. That means full flexibility when workflows outgrow visual builders.
I have a few slots open for the right long-term partnership. Happy to walk through my architecture approach on a call – DM me here or email: priyanshukumarmaurya2224@gmail.com