4 local tools in this family
AI Prep Tools
Tools built for the AI workflow: repair broken JSON from chatbots, mask sensitive details before pasting, estimate exact vs. approximated token budgets, and inspect documents for hidden or invisible text.
LLM JSON Repair & Schema Check
Repair malformed model JSON, inspect every change with per-rule undo, and validate schemas.
Safe-to-Paste Redactor
Mask emails, phones, cards, IBANs, IPs, and API keys locally — then restore originals in AI replies.
Token & Context Calculator
Honest Exact vs. Estimated token counts, context-fit meter, dated API pricing, and chunk splitter.
Hidden Text & Invisible Character Scanner
Inspect text, HTML, DOCX, and PDFs for zero-width chars, Unicode tags, white/tiny text, and hidden layers.
How to Choose the Right Tool
Start with Safe-to-Paste Redactor before sharing work documents with a chat model, check context limits in Token & Context Calculator, verify incoming files with Hidden Text Scanner, and fix malformed model responses in LLM JSON Repair.
Related Guides
Trust & Verification
How to Check in Your Browser That a Tool Never Uploads Your Files
Do not take any website's privacy badge on faith. Follow this 60-second DevTools Network tab check to confirm every byte stays on your machine.
AI Prep
What a Pattern-Based PII Redactor Can and Cannot Find
Understand why deterministic checksums catch cards, IBANs, emails, and API tokens reliably — and where human review remains essential.
AI Prep & Dev
Why LLM 'JSON Mode' Still Breaks — and How Deterministic Repair Fixes It
A practical breakdown of markdown fence wrapping, trailing commas, Python literals, and mid-stream truncation in model responses.
AI Prep & Security
Invisible Prompt Injections & Hidden Unicode: What Can Hide Inside Documents
How zero-width characters, Unicode tag blocks, white-on-white PDF text, and tiny fonts conceal instructions inside everyday files.
AI Prep
Why Token Counts Differ Across OpenAI, Claude, and Gemini
How BPE vocabularies diverge across English prose, source code, JSON, and non-Latin text — and why honest estimates beat false precision.