ai-analyst-lab/ai-analyst

What it solves

This project provides an AI-powered data analyst that integrates directly into Claude Code. It transforms the process of data analysis from simply generating SQL queries to performing a complete analytical workflow—including data profiling, validation, and the creation of professional deliverables like briefs and slide decks.

How it works

The system is built on three architectural layers:

• Skills: Automated standards (defined in .claude/skills/) that ensure the analyst follows best practices, such as styling charts correctly or providing comparisons for every number reported.
• Pipeline Agents: Multi-step workflow definitions (located in agents/) that handle complex, end-to-end jobs like full investigations or forecasting.
• Helpers: Deterministic Python libraries that handle the actual math and statistics, including causal inference, experiment design (A/B testing), and forecasting, ensuring that the LLM reasons while the code performs the arithmetic.

It maintains a .knowledge/ folder as a persistent memory store for schema notes, metric definitions, and a log of user corrections to prevent repeating mistakes.

Who it’s for

Data analysts, business intelligence professionals, and decision-makers who want to automate the tedious parts of data exploration and analysis while maintaining high standards of rigor and trust.

Highlights

• Broad Data Connectivity: Supports CSVs, DuckDB, Postgres, BigQuery, Snowflake, Databricks, Redshift, SQL Server, and MySQL.
• Rigorous Validation: Includes an evaluation control plane to measure reliability, trace findings back to source rows, and triangulate results.
• Professional Deliverables: Can export findings to Google Docs, Slides, Notion, PDF, Word, Slack, or email.
• Advanced Analytics: Built-in support for A/B test design, power analysis, and causal inference (e.g., difference-in-differences).
• Persistent Knowledge: Remembers business glossaries and user-corrected errors across sessions.