dataclaw
Agent harness to publish your agent chat history as Huggingface datasets.
DataClaw is an AI agent harness that converts coding agent conversation histories into structured Hugging Face datasets. It parses session logs from Claude Code, Codex, and other AI coding assistants, automatically redacts secrets and personally identifiable information, and publishes the sanitized data for research and model training purposes. The tool supports both a macOS menu-bar application and a command-line interface, guiding agents through a six-step export workflow that includes privacy review and user attestation before publication. All exported datasets are tagged with 'dataclaw' on Hugging Face, contributing to a distributed collection of real-world human-AI coding collaboration examples.
Key Features
Use Cases
- 01Creating training datasets from real-world human-AI coding sessions for model fine-tuning
- 02Building open collections of agent conversation examples for research and analysis
- 03Archiving and sharing coding agent session histories while protecting sensitive information
- 04Contributing to distributed datasets of human-AI collaboration patterns
- 05Analyzing tool usage patterns and token consumption across multiple coding projects
- 06Exporting conversation histories from Claude Code, Codex, and similar coding assistants for backup or portability
Related Agents
View morehermes-agent
The agent that grows with you
agency-agents
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
openinterpreter
A coding agent for open models like Kimi K3
cline
Autonomous coding agent as an SDK, IDE extension, or CLI assistant.
dataclaw — FAQ
What is DataClaw?+
DataClaw is an agent harness that exports your AI coding assistant conversation histories (from Claude Code, Codex, etc.) as structured datasets and publishes them to Hugging Face. It automatically redacts secrets, API keys, and personally identifiable information before export.
How do I install DataClaw?+
For Apple Silicon Macs, download the DMG from the GitHub releases page and drag DataClaw.app to Applications. For CLI usage or other platforms, install via pip with 'pip install -U dataclaw'. The Mac app is currently unsigned so requires a one-time Gatekeeper approval.
Which AI coding agents does DataClaw work with?+
DataClaw currently supports Claude Code and Codex, with agent skill integration available for Claude Code. It can be extended to support other coding agents by implementing new parsers.
Do I need a Hugging Face API key?+
Yes, you need a Hugging Face account and token to publish datasets. When automating with an agent, use the '--token' flag rather than running 'hf auth login' interactively.
Is DataClaw free and open source?+
Yes, DataClaw is open source and released under the MIT license. The tool is free to use, though you'll need your own Hugging Face account to publish datasets.
How does DataClaw protect my privacy?+
DataClaw applies username redaction, regex-based secret detection (API keys, tokens, passwords), entropy analysis for unusual strings, email redaction, and custom redaction rules. However, it recommends manual review with tools like trufflehog before publishing, as automated redaction cannot catch everything.
How do I install dataclaw?+
Open the source repository on GitHub and follow its README. dataclaw is a agent — MCP Agents Market links you directly to the official repo.
Is dataclaw free?+
dataclaw is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.