/tools/log-redactor

Log & Diagnostic Data Redactor

Remove likely secrets, tokens, cookies and personal data before sharing diagnostics with AI assistants, support teams or public issues.

diagnostic-redactor://local

1. Diagnostic source

Choose one log, JSONL or HAR file, or paste a small diagnostic sample up to 10 MB.

Processed locally · source remains unchanged

2. Scope and evidence

Load a source to filter
Level
Advanced evidence filters

Evidence filters

Likely exposure

3. Records

4. Local details

Exposure findings

No sampled signals

The sharing-safe policy removes base64 and binary HAR body content from the derivative.

5. Protection policy

Create a sanitised derivative

Every selected rule is applied to a new output and the transformed records are scanned again before release.

Protect
Protect something specific
  • No custom session rules.

Custom rules stay in this tab only and are omitted from reports and analytics.

The sharing-safe preset protects every built-in category and preserves useful correlations. Preview remains bounded; export verifies the complete selected scope.

Choose diagnostic data or load the synthetic sample. Processing stays in this browser.

/before-sharing

Preserve debugging value without quietly exposing source data

Context-aware replacements

Repeated email addresses, IPs and tokens receive stable session labels, so relationships remain visible without retaining the original value.

Structured where possible

JSONL and HAR values are changed by path so the derivative remains valid structured data. Unchanged records retain their original bytes.

Coverage is explicit

Oversized or otherwise uninspectable content cannot receive a verified verdict. Review-required output needs a separate acknowledgement.

Your own sensitive terms

Add exact values or field names for internal project, tenant or customer identifiers that generic detectors cannot know.

/limitations

What deterministic redaction can and cannot establish

The tool looks for defined credential, token, cookie, payment, email, phone, IP, path and custom signals. Ordinary names, postal addresses, proprietary source code and every future credential format cannot be inferred reliably without more context. Review important derivatives before distributing them.

/faq

Log redaction questions

Are logs uploaded?

No. Source inspection, transformation, verification and report generation run in local browser workers.

Can I use the result with ChatGPT or another AI?

Yes. Copy a small verified derivative or download a larger file, review it, and provide it to the service you choose. Data Demon does not send it automatically.

What does verified mean?

No selected deterministic rule matched the transformed output and no inspection coverage gap was recorded. It is not a universal guarantee.

Why are binary HAR bodies removed?

Encoded binary payloads cannot be assessed reliably with text detectors. Removing their content gives the sharing-safe policy a clearer privacy boundary while retaining MIME and size metadata.