A Browser Tool Scores How Much Your Setup Looks Like a Chinese User to Claude Code
Regional restrictions on AI coding tools are increasingly enforced through client-side environment fingerprinting rather than just IP geolocation. This tool makes those signals explicit and testable, so developers who travel, use VPNs, or run internationalized systems can see exactly which local settings might lock them out.
FuckClaude runs entirely in the browser and turns six fingerprintable characteristics into a weighted scoring model totaling 100 points. System timezone carries the most weight at 30 points, followed by browser language at 24 and installed Chinese fonts at 20. Intl locale settings, UTC+8 offset, and OS-level emoji rendering style round out the remaining 28 points. Each signal gets a 0-to-1 similarity score, multiplied by its weight and summed into a final Low, Medium, or High verdict.
The tool reverse-engineers the signals Claude Code likely uses for regional classification, surfacing exactly which browser and system properties contribute to a match. Canvas-based font detection checks for Microsoft YaHei and PingFang SC, while the emoji check infers the system vendor from the User-Agent string.
Source code is available on GitHub under the name FuckClaude, making the detection logic fully auditable and forkable.
Client-side fingerprinting for regional enforcement is coarse and brittle — changing a single setting like browser language or timezone can swing the score dramatically, which means the classification is easy to evade but also prone to false positives for legitimate international users.
Publishing the scoring weights and detection methods as open-source code turns a black-box restriction into a transparent, auditable checklist, which shifts the power dynamic back toward the user.