alice/ai: Multi-Agent AI Assistant SaaS (Client Project)
Tran Tien Van
Client project. Built by Van Data Team for alice/ai (formerly PraxisAI), an AI operating system for small business. Live product: goalice.ai
The challenge
Small-business owners lose their week to admin spread across the tools they run the business on, and an assistant that can only answer questions cannot do that work.
What we built
A LangGraph Supervisor that routes each request to specialised sub-agents across 70+ MCP tool integrations
A FastAPI backend of 51 specialised controllers on AWS ECS Fargate, with retrieval on Pinecone and FAISS and tracing in LangSmith
GPT-4 and Gemini behind the agents; answers streamed over server-sent events
A React 19 + Vite frontend shipped as a PWA and, through Capacitor 7, as native iOS and Android apps
AWS Cognito multi-tenant sign-in, Stripe billing and 14 DynamoDB tables
MCP connections to Cliniko and Nookal for healthcare practice records and scheduling
Results
70+ MCP tool integrations under one LangGraph Supervisor
51 specialised backend controllers on FastAPI and AWS ECS Fargate
One React 19 codebase shipping web, PWA, iOS and Android
Key decision
The supervisor routes to specialised sub-agents instead of handing one agent every tool, so each agent chooses from a short, relevant toolset even as integrations pass 70.
Stack: LangGraph, LangSmith, MCP, Pinecone, FAISS, GPT-4, Gemini, FastAPI, AWS ECS Fargate, React 19, Capacitor, AWS Cognito, DynamoDB, Stripe, GitHub Actions.
Full case study: vandatateam.com/portfolio
Like this project
Posted Nov 5, 2026
Client project for alice/ai: LangGraph supervisor over 70+ MCP tool integrations, 51 FastAPI controllers on AWS, one React 19 app for web, iOS and Android.
