AI Data Inspector: analyze your AI chat history without ever uploading it.

What the tool does

Drop a ZIP, JSON or JSONL export onto the page and the analyzer immediately reads it — decompressing archives in your browser's memory with JSZip. From there you get a complete view of your history:

• Full-text search across every thread, so you can find that one prompt from months ago.
• PII & secret scanning that flags API keys, passwords, tokens, emails and credit-card-like patterns.
• Custom-instruction isolation, pulling out the system prompts you configured in each platform.
• Activity heatmaps showing when and how much you talk to each assistant.
• Cost estimates and clean Markdown / text / JSON exports.

Why local-first matters

Cloud tools that "sync" your history route your prompts through their servers — which means confidential code, business strategy and personal data pass through infrastructure you don't control. AI Data Inspector uses the opposite architecture. Decompression, parsing, search and scanning all run in a Web Worker on your own CPU. There is no upload endpoint, no analytics beacon, and no telemetry on the analyzer itself. Close the tab and every trace is gone from memory.

Which exports are supported

The tool speaks the native formats of the major assistants: ChatGPT's conversations.json, Claude's chat_messages.jsonl, Gemini's Gemini_History.json, DeepSeek chat maps, plus role-based JSON and JSONL from local models and other assistants. A universal parser handles anything shaped like { role, content }, including Ollama, Open WebUI and ShareGPT-style exports.

Get started

Export your history from the platform's settings, then drop the file straight onto the analyzer. No account, no sign-up, no install — and it keeps working offline. Use it to find answers, audit what you've shared, rotate leaked credentials, back up your history, or migrate between assistants without losing your best threads.