dataclaw 0.1.0

Export your Claude Code conversations to Hugging Face as structured training data

DataClaw

Turn your Claude Code conversation history into structured training data and publish it to Hugging Face with a single command. DataClaw parses session logs, redacts secrets and PII, and uploads the result as a ready-to-use dataset — so the hours you spend coding with Claude can help make future models better.

All DataClaw datasets are tagged dataclaw — browse them all.

Getting started

Paste this into Claude Code (or any coding agent):

Help me export my Claude Code conversation history to Hugging Face using DataClaw.
Install it, set up the skill, then walk me through the process.

STEP 1 — INSTALL
  pip install dataclaw
  If that fails: git clone https://github.com/banodoco/dataclaw.git /tmp/dataclaw && pip install /tmp/dataclaw
  If that also fails, ask the user where the source is.

STEP 2 — INSTALL SKILL
  dataclaw update-skill claude

STEP 3 — START
  dataclaw prep
  Every dataclaw command outputs next_steps in its JSON — follow them through the entire flow.

IMPORTANT: Never run bare `huggingface-cli login` — always use --token.
IMPORTANT: Always export with --no-push first and review for PII before publishing.

Manual usage (without an agent)

Quick start

pip install dataclaw
huggingface-cli login --token YOUR_TOKEN

# See your projects
dataclaw prep

# Configure
dataclaw config --repo username/dataclaw-username
dataclaw config --exclude "personal-stuff,scratch"
dataclaw config --redact-usernames "my_github_handle,my_discord_name"
dataclaw config --redact "my-domain.com,my-secret-project"

# Export locally first
dataclaw export --no-push

# Review the JSONL, then push
dataclaw export

Commands

Command	Description
dataclaw prep	Discover projects, check HF auth, output JSON (read-only)
dataclaw list	List all projects with exclusion status
dataclaw config	Show current config
dataclaw config --repo user/dataclaw-user	Set HF repo
dataclaw config --exclude "a,b"	Add excluded projects (appends)
dataclaw config --redact "str1,str2"	Add strings to always redact (appends)
dataclaw config --redact-usernames "u1,u2"	Add usernames to anonymize (appends)
dataclaw export	Export and push
dataclaw export --no-push	Export locally only (review first)
dataclaw export --all-projects	Include everything (ignore exclusions)
dataclaw export --no-thinking	Exclude extended thinking blocks
dataclaw update-skill claude	Install/update the dataclaw skill for Claude Code

What gets exported

Data	Included	Notes
User messages	Yes	Full text (including voice transcripts)
Assistant responses	Yes	Full text output
Extended thinking	Yes	Claude's reasoning (opt out with --no-thinking)
Tool calls	Yes	Tool name + summarized input
Tool results	No	Not stored in Claude Code's logs
Token usage	Yes	Input/output tokens per session
Model & metadata	Yes	Model name, git branch, timestamps

Privacy & Redaction

DataClaw applies multiple layers of protection:

1. Path anonymization — File paths stripped to project-relative
2. Username hashing — Your macOS username + any configured usernames replaced with stable hashes
3. Secret detection — Regex patterns catch JWT tokens, API keys (Anthropic, OpenAI, HF, GitHub, AWS, etc.), database passwords, private keys, Discord webhooks, and more
4. Entropy analysis — Long high-entropy strings in quotes are flagged as potential secrets
5. Email redaction — Personal email addresses removed
6. Custom redaction — You can configure additional strings and usernames to redact
7. Tool input pre-redaction — Secrets in tool inputs are redacted BEFORE truncation to prevent partial leaks

This is NOT foolproof. Always review your exported data before publishing. Automated redaction cannot catch everything — especially service-specific identifiers, third-party PII, or secrets in unusual formats.

To help improve redaction, report issues: https://github.com/banodoco/dataclaw/issues

Data schema

Each line in conversations.jsonl is one session:

{
  "session_id": "abc-123",
  "project": "my-project",
  "model": "claude-opus-4-6",
  "git_branch": "main",
  "start_time": "2025-06-15T10:00:00+00:00",
  "end_time": "2025-06-15T10:30:00+00:00",
  "messages": [
    {"role": "user", "content": "Fix the login bug", "timestamp": "..."},
    {
      "role": "assistant",
      "content": "I'll investigate the login flow.",
      "thinking": "The user wants me to look at...",
      "tool_uses": [{"tool": "Read", "input": "src/auth.py"}],
      "timestamp": "..."
    }
  ],
  "stats": {
    "user_messages": 5, "assistant_messages": 8,
    "tool_uses": 20, "input_tokens": 50000, "output_tokens": 3000
  }
}

Each HF repo also includes a metadata.json with aggregate stats.

Finding datasets on Hugging Face

All repos are named {username}/dataclaw-{username} and tagged dataclaw.

• Browse all: huggingface.co/datasets?other=dataclaw
• Load one:

  from datasets import load_dataset
  ds = load_dataset("alice/dataclaw-alice", split="train")
  

• Combine several:

  from datasets import load_dataset, concatenate_datasets
  repos = ["alice/dataclaw-alice", "bob/dataclaw-bob"]
  ds = concatenate_datasets([load_dataset(r, split="train") for r in repos])
  

The auto-generated HF README includes:

• Model distribution (which Claude models, how many sessions each)
• Total token counts
• Project count
• Last updated timestamp

Code Quality

License

MIT