Systems / Nora AI Agent System
Agent and workflow architecture built for real delivery.
> stack: n8n / Claude API / Supabase / Vercel
◆ SYSTEM BRIEF
> what_problem_does_this_solve?
Most AI agent demos break outside of controlled conditions. Nora was built to test what a real delivery-ready agent system looks like — with proper intent routing, tool selection, memory, and graceful failure handling.
◆ ARCHITECTURE
◆ SYSTEM FLOW
User request + context
Tool scope, token limits
Claude API, n8n, Supabase
Session state + vector store
Agent response + action log
◆ STACK
AI & Agents
Claude API
n8n
Data & Infra
Supabase
Vercel
Design & Interface
Figma
SVG
◆ OUTCOME
> what_does_it_do_now?
Nora routes intent, selects tools, manages memory across sessions, and handles edge cases without breaking. It runs on n8n with Claude API as the reasoning layer, Supabase for persistence, and Vercel for deployment. Every layer is wired — there are no gaps between the demo and the deployment.
◆ RELATED SYSTEMS