Our in-house analytics agent & Lessons learned from agent-to-agent communication

A walkthrough of building a self‑serve BI tool using Streamlit on EC2 with Claude‑generated queries, followed by real‑world lessons from three months of agent‑to‑agent communication.

Overview

This is a rapid, two-part “show-and-tell” straight from our build logs.

The 1-Day BI Platform: We’ll show how we built a self-serve internal BI tool for our (very small) team. We’ll walk through the implementation of a simple Streamlit app, hosted on EC2, that uses Claude to generate and run data queries on the fly. This lets our non-technical folks “vibe up” dashboards instantly. We’ll show the core code and prompt structure.

When Agents Talk to Agents: We’ll share technical insights from having our two core AI agents, “Jack” and “Jill,” in production and communicating with each other for the last three months. This isn’t a product pitch; it’s a technical review of the architecture and the messy discoveries we made building a product reliant on agent-to-agent communication.

Tech stack