ClaudeDown is a real-time complaint tracking tool for Claude AI users. The core question it aims to answer is: when users feel that Claude's performance has gotten worse, is it their own problem, or are many people encountering the same situation at the same time?
In terms of purpose, the website focuses on monitoring complaint signals about Claude AI on public platforms. It scrapes tweets mentioning Claude on X (formerly Twitter), identifies complaint-type content from them, and uses "the proportion of complaint tweets among all tweets mentioning Claude" as the core metric. The higher the proportion, the more people are complaining.
Its main capabilities include: first, status judgment; the website gives a conclusion similar to "Is Claude getting dumber right now?", for example showing "all normal"; second, a complaint timeline, showing changes in the proportion of complaints over time; third, a time-period comparison view, comparing the complaint level of each time slot with the historical median for that period, using color to indicate the relative level, with 1.0x as the typical level and above 2.0x regarded as elevated; hovering over or clicking a cell shows detailed data, including the complaint rate and relative level; fourth, recent tweet samples, displaying recent tweets about Claude AI sorted by engagement; fifth, spike alerts, which users can subscribe to and receive notifications only when an anomalous spike occurs. The website also displays the Claude Code version number (v 2.1.296 in the material) and provides AI-generated summaries.
In terms of usage, users can directly visit the website to view the current status, complaint timeline, and time-period comparison chart, with no installation required. If they need to receive anomaly alerts, they can leave contact information through the subscription entry provided on the website. All times are labeled as UTC.
The main applicable scenarios are: when developers or users suspect that Claude has experienced performance degradation, using it to cross-verify whether it is a widespread phenomenon; when discussing model stability within a team, using it as a reference for public complaint signals; and it can also be used to observe whether complaints are concentrated and elevated during a specific period.
In terms of limitations, the official statement makes clear: this is a community project, the data is based on sampled tweets, and it is an estimate rather than full data; both complaint volume and baseline are estimated values; spikes are detected by an algorithm and may include false positives. Therefore, the website's results should not be treated as an official status page or authoritative outage report, but only as an auxiliary reference for public complaint signals.
The website was built by Prabal, Saransh, and Sarthak, and the official website provides links to published articles. Apart from the above content, the material does not provide pricing methods, specific subscription prices, supported platforms, or a more detailed technical implementation description.
