AI Analyst is a Claude Code-powered toolkit designed to automate the time-consuming aspects of product analysis, transforming business questions into validated slide decks with speaker notes in minutes. It targets analysts by handling approximately 80% of their typical workload, enabling them to focus on validation and higher-level strategic thinking. The primary benefit is a significant acceleration of the analysis and reporting cycle, allowing for more frequent and comprehensive insights.

How It Works

The core of AI Analyst is a sophisticated pipeline orchestrated by a DAG (Directed Acyclic Graph) engine, enabling parallel execution of 18 specialized agents. These agents operate across four phases: Framing the question, Analyzing the data, Building the narrative, and Creating the slide deck. Version 2 introduces a persistent knowledge system that captures user corrections, query patterns, and business context, allowing the tool to self-learn and avoid repeating mistakes. It supports a variety of data sources, including CSV, DuckDB, Postgres, BigQuery, and Snowflake, and is dataset-agnostic.

Quick Start & Requirements

1. Install Claude Code: npm install -g @anthropic-ai/claude-code (Requires a Claude Pro subscription).
2. Clone and Setup:

   git clone https://github.com/ai-analyst-lab/ai-analyst.git
   cd ai-analyst
   pip install -e ".[dev]"
   

3. Start Claude Code: claude
4. Connect Data: Use /connect-data or directly run /run-pipeline data_path=data/my_csvs/ question="Why is conversion dropping?".

• Prerequisites: Python 3.10+, Node.js 18+, Claude Pro subscription.
• Documentation: Setup guide (docs/setup-guide.md), Theming (docs/theming.md).

Highlighted Details

• Features 18 specialized agents and 39 auto-applied skills.
• Employs a DAG engine for efficient parallel agent execution.
• Supports direct data connections to CSV, DuckDB, Postgres, BigQuery, and Snowflake.
• Adheres to the "Storytelling with Data" methodology for chart design and presentation.
• Includes a self-learning knowledge system that persists corrections and context.
• Automated validation, root cause analysis, and narrative generation.
• Exports include PDF and HTML slide decks with speaker notes.

Maintenance & Community

For assistance or to report bugs, users are directed to open a GitHub Issue. Specific community channels like Discord or Slack are not detailed in the README.

Licensing & Compatibility

The project is released under the MIT License, allowing for broad usage and modification.

Limitations & Caveats

This tool is designed as an assistant for expert analysts, not a replacement. It requires user validation of its output, as it may misinterpret metrics or select incorrect data columns without expert oversight. It is a starting point and requires user input to adapt to specific business contexts and data. It does not work out-of-the-box and relies on user corrections to grow and improve accuracy for a given use case.