What is claudescraper?

This project reads the usage numbers shown inside Claude Code's interactive terminal panel and turns them into a web dashboard and a JSON endpoint. The tool exists because Claude Code does not offer a scriptable way to check account wide plan usage, only single query cost or plain login status, so this instead drives the actual terminal interface in the background and reads what appears on screen, on a repeating timer, then caches the result. It runs as a single always on Docker container. On first run it walks a user through logging into Claude Code directly from the web page, no terminal commands needed, and after that credentials persist in a named volume so restarts go straight back to the dashboard. A user can also skip the web login entirely by supplying an existing token or an Anthropic API key as an environment variable. The dashboard shows progress bars for things like the current session and weekly usage windows, along with reset times and cost estimates, refreshed automatically every five minutes by default, with a button to force an immediate refresh. If a background refresh fails, the dashboard keeps showing the last successful data rather than an error, and simply marks it as stale. A settings page lets a user record which subscription plan they are on, since Claude Code itself does not expose that, and includes an optional toggle to automatically send a test prompt whenever a usage window resets. The underlying scraper can also be run as a one time command line tool outside the web server, returning either a human readable table or structured JSON, which is useful for scripting or checking usage from other tools. The project explicitly has no built in access control, so it is meant to be run behind a trusted network such as localhost, a private network, or your own reverse proxy rather than exposed directly to the internet.