What is scrapemychats?

scrapemychats solves a specific problem: when someone leaves a ChatGPT business or Team account, there is often no built-in way to export conversation history. The personal account "Export data" button is simply missing from these workplace plans. This tool copies every conversation, including text and attached files, onto your own computer and then builds a searchable offline archive you can keep indefinitely. It also works on personal accounts. The tool operates by opening a real Chrome browser window on your machine using your existing ChatGPT login session. It never sees your password, and no data is sent anywhere except to chatgpt.com itself. Everything stays local. The export process runs unattended for a few hours on large accounts, which is intentional, and if your computer sleeps or you interrupt it, you just run the same command again and it picks up where it left off. The end result is a folder containing a separate subfolder for each conversation. Inside each one you get the complete raw data in JSON format, a readable Markdown transcript, and any attached files or images. A manifest CSV file tracks the status of every chat, and an errors log records anything that could not be fetched. The centerpiece is a single self-contained HTML file called viewer.html that lets you search and browse everything at once. That viewer requires no server, no internet connection, and no additional software. You can run it from a USB stick. It offers instant full-text search across all conversations with highlighted matches, optional categorization into groups like Personal and Work based on keywords you define, inline image thumbnails, and keyboard navigation. You can edit a categories file to match the structure of your own life and re-run the build step to re-file everything. The README is unusually thorough for non-technical users, providing step-by-step instructions for Windows and Mac with about ten steps each, and even suggesting you paste a prompt into a chatbot to get walked through the process one instruction at a time. The full README is longer than what was shown.