AgentGraph Kit

I have been working as an AI Engineer for the last few years, and I noticed a frustrating cycle in every new project I started.

Whenever I had an idea for a new Agent or SaaS, the excitement would wear off quickly. Why? Because instead of coding the actual AI logic, I was stuck setting up the "boring" plumbing.

The Real Problem: The Python-JS Gap

We all know the dilemma:

• We need Next.js for the frontend (UI, Auth, Payments).

• We need Python for the actual AI Agents (LangGraph, Heavy lifting).

Connecting these two worlds is surprisingly hard. You have to handle authentication across two different services, set up asynchronous streaming so the UI doesn't freeze while the agent thinks, and manage the shared state.

I found myself rewriting the same "glue code" over and over again to make them talk to each other securely.

The Solution

Last month, I decided to stop building features and start building a system. I took my internal production stack and cleaned it up into a reusable boilerplate.

I call it AgentGraph Kit.

It is basically a pre-wired starter kit that solves the integration headache out of the box.

• Frontend: Next.js 14 (App Router) fully configured.

• Backend: FastAPI setup specifically for LangGraph agents.

• The Glue: It handles the API handshake, streaming, and database connections (Supabase) so you don't have to.

I just launched it today. My goal is to save other developers those first 40 hours of setup so you can get straight to defining your Agent's logic.

If you are interested, you can check the architecture here: https://agentgraphkit.com

I also added a code (AGENT50) for the Indie Hackers community to get it for $99 instead of the full price.

Let me know if you have any questions about the structure.

Best, Mario