kruzovic7/ai-data-extractor

What it solves

This tool allows users to extract their local chat history from various AI coding assistants (such as Cursor, Claude Code, Aider, and Windsurf) into a single, normalized JSONL format. This is useful for backing up conversations, performing personal analytics, or preparing datasets for fine-tuning AI models.

How it works

The toolkit detects the user's operating system and searches common application data directories for known storage formats. It reads data from various sources—including SQLite databases, JSON files, and Markdown transcripts—using read-only connections to avoid blocking running applications. It then normalizes this disparate data into a consistent schema containing messages, source, and session IDs.

Who it’s for

Developers who use multiple AI coding assistants and want to own their data, as well as ML engineers who wish to use their own interaction history as training data for fine-tuning.

Highlights

• Broad Support: Supports a wide range of tools including Cursor, Claude Code, Codex CLI, Windsurf, Trae, Continue, Gemini CLI, OpenCode, Cline/Roo Code, and Aider.
• Cross-Platform: Automatically handles search paths for macOS, Linux, and Windows.
• Zero Dependencies: Built using only the Python standard library.
• Flexible Extraction: Supports interactive menus, direct CLI flags, and the ability to merge all extracted conversations into one file.
• Extensible: Provides a simple two-function interface for adding new data sources.