**Anthropic** confirmed that the source code for **Claude Code**, its closed-source AI coding assistant, was accidentally leaked. According to the company, no customer data or credentials were exposed during the incident. ### The Accidental Leak While **Anthropic** actively supports the open-source community, **Claude Code** has remained a proprietary offering. However, a recent update inadvertently included internal source code. In a statement to BleepingComputer, **Anthropic** stated: "Earlier today, a **Claude Code** release included some internal source code. No sensitive customer data or credentials were involved or exposed. This was a release packaging issue caused by human error, not a security breach. We're rolling out measures to prevent this from happening again." The leak was initially spotted by Chaofan Shou (@Fried_rice), and the code quickly proliferated across **GitHub** and other platforms.  The source code was mistakenly included in **Claude Code** version 2.1.88, which was briefly published on NPM. This version contained a 60 MB file named `cli.js.map` that held the entire source code for the latest version. A source map file is typically used for debugging, linking compiled JavaScript back to its original source code. When these map files include a "sourcesContent" field with the full text of the original source files, the entire source code tree can be reconstructed. ### Source Code Details The reconstructed source code comprises approximately 1,900 files, totaling 500,000 lines of code. It also reveals details about several **Claude**-exclusive features. **Anthropic** has initiated **DMCA** infringement notifications to remove the leaked code from various online repositories.  Developers have already begun analyzing the leaked code to uncover undocumented features and understand the application's inner workings. Reports indicate that **Anthropic** is testing a new "Proactive mode," where **Claude** will code continuously. Additionally, a "Dream" mode has been identified, where **Claude** can think in the background, develop ideas, and solve problems autonomously. ### Usage Limit Bug In separate news, users have reported that **Claude** has quietly reduced usage limits. Users on both the Pro and Max plans are experiencing faster exhaustion of their allotted usage. This behavior has been observed across various accounts, with usage rapidly increasing after only a few interactions with **Claude Code**. **Anthropic** has acknowledged the issue and confirmed that they are investigating a bug that causes usage limits to deplete more quickly. According to Lydia Hallie from **Anthropic**, the team is actively working to resolve the issue, prioritizing it as a top concern. As of March 31, the issue remained unresolved, with **Anthropic** providing ongoing updates. Some users speculate that the change might be intentional, given **Claude**'s increasing popularity, but this remains unconfirmed.