Ai Analyst Genome

AI Analyst Genome

[!IMPORTANT] **Note from Shane — February 22, 2026**

This repo was last pushed to on February 19. Since then, the system has grown significantly through testing on more complex, real-world datasets. Everything below works, but the version we're running locally is substantially ahead of what's here. Major update coming in the next couple of days.

**A self-building AI Data Analyst.** Drop this into an empty repo with Claude Code, answer 7 questions, and it builds itself — agents, skills, helpers, templates, and a full analysis pipeline.

Powered by [AI Analyst Lab](https://aianalystlab.ai) | Created by Shane Butler, Sravya Madipalli, and Hai Guan

---

What This Is

The AI Analyst Genome is a single document that bootstraps a complete AI Data Analyst product through a 3-layer inception model:

1. You answer 7 setup questions — where's your data, what do you want to analyze, who's the audience
2. 4 architect agents debate and produce a build plan
3. 7 builder agents execute the plan, creating 19 analysis agents, ~30 skills, and 6 Python helper modules
4. You have a working AI Data Analyst that can take a business question from framing through validated analysis to a finished slide deck

Requirements

• Claude Code running on Amazon SageMaker via Amazon Bedrock (or any Claude Code environment)
• Recommended model: Claude Opus 4.6
• An empty repo or directory
• Access to your data platform (Snowflake, PostgreSQL, BigQuery, CSV files, or anything else)

Quick Start

# 1. Clone this repo (or copy the two files into an empty directory)
git clone https://github.com/ai-analyst-lab/ai-analyst-genome.git
cd ai-analyst-genome

# 2. Open Claude Code
claude

# 3. That's it. Claude reads CLAUDE.md, reads the genome, and starts
#    the setup wizard automatically. Answer 7 questions and the build begins.

What Gets Built

| Component | Count | Description | |------