On the evening of May 26, Moonshot AI's Kimi Code team reached an agreement with the open-source community, ending a crisis sparked by account suspensions, subscription risk controls, and accusations of double-standard PR. Open-source developer Leechael announced they would delete earlier inflammatory posts and confirmed the pi-provider-kimi-code plugin will continue to be maintained. Paid subscribers can still safely use Kimi Code Plan through third-party tools like Pi Coding Agent.
https://twitter.com/Leechael/status/2059202023539822987
Kimi Code team member Young published an official Chinese statement explaining the technical details behind the account blocks. To combat token reselling and suspicious requests, the team had upgraded its risk-control defenses, using User-Agent strings as one detection criterion. However, the model failed to account for the fact that many legitimate third-party Code Agents (such as Pi) mimic Kimi CLI's User-Agent, leading the system to flag normal requests as malicious and mistakenly blocking numerous compliant subscribers.
The double-standard controversy turned out to be a timing mismatch across communication channels. Young explained to this channel that because most domestic LLM developers gather in Moonshot's official Feishu (a workplace platform) group, the team addressed concerns there first. The delayed English reply on X/Twitter — due to time zone differences and cross-platform operations — inadvertently created the perception that apologies were only offered to the English-speaking audience. Moonshot AI has since fully adjusted its risk-control strategy, restored all affected accounts, and proactively contacted third-party developers to refine User-Agent detection.
After the settlement, Leechael said Kimi's research team had an in-depth discussion with them, rebuilding basic confidence in the domestic LLM ecosystem, and has started writing exclusive prefix caching optimization guides for Kimi's personal agents.