dataclaw

Why we included this project

DataClaw takes the session logs that coding agents like Claude Code and Codex leave behind and turns them into structured datasets you can share. It parses the conversation history, redacts secrets and personally identifiable information, then uploads the result to Hugging Face in one command, with every export tagged so it is easy to find. For teams training their own models or fine-tuning on real human-AI coding collaboration, this is a practical way to gather training data you actually own rather than relying on vendor-provided data. It runs as a menu-bar app on Apple Silicon Macs or as a CLI, so you do not need to set up a Python environment just to export your own agent history. The redaction and the review-before-publish step are worth knowing about, since the tool itself admits automated scrubbing is not foolproof.