Anthropic accidentally leaks Claude Code source code

Anthropic has accidentally published the complete source code for Claude Code, its popular AI coding tool, via the npm registry. The leak occurred on March 31, 2026, when version 2.1.88 was published with a 60 MB source map file containing the full source code.

Security researcher Chaofan Shou first spotted the error, and the code quickly spread across GitHub. The reconstructed codebase comprises approximately 1,900 files and 500,000 lines of TypeScript code, revealing internal architecture, memory management, orchestration logic, and 44 hidden features.

Anthropic confirmed this was a packaging error caused by human mistake, not a security breach. No customer data or credentials were exposed. This marks at least the second time the company has experienced such a source code exposure.

The leak also revealed details about an upcoming model called "Claude Mythos" and a new model tier called "Capybara", expected to surpass the current Opus tier in intelligence.