← Back to stories Detailed view of colorful programming code on a computer screen.
Photo by Markus Spiske on Pexels
IT之家 2026-04-01

Claude Code Just Open‑Sourced: 510,000 Lines of Code, Internet Celebrates

Leak overview

It has been reported that Anthropic’s Claude Code was effectively exposed after a source‑map file on the npm registry leaked the product’s compiled-to-source mapping. According to coverage by IT Home (IT之家) and screenshots circulating on X, the public package included a 59.8 MB cli.js.map that reportedly maps back to more than 1,900 files and about 512,000 lines of TypeScript. Developers on social platforms quickly pulled and inspected the files; the moment spread across global developer communities.

What the files reveal

Reportedly, the dumped material contains the full Claude Code CLI implementation: a React + Ink terminal UI running on the Bun runtime, a roughly 46,000‑line QueryEngine.ts that handles token counting and chain‑of‑thought loops, and a “Tools” suite of 40+ modules (file I/O, Bash execution, LSP integration, sub‑agent generation). The leak also shows multi‑agent infrastructure — a coordinator and a bridge to VS Code/JetBrains — along with persistent background sessions, memory integration, and an autonomous daemon design. Curiosities include an embedded “Buddy System” (a Tamagotchi‑like pet set) and a so‑called “Undercover Mode” that reportedly strips AI traces from public commits when Anthropic employees push to public repos.

Why it matters

This is more than an engineering scoop. If accurate, the exposure hands outsiders detailed implementation logic, internal APIs and automated decision flows that companies normally keep proprietary. That raises intellectual‑property and security risks and could complicate compliance with export controls and platform policies — a notable point as Western AI firms already operate under intensified regulatory and geopolitical scrutiny. There was no immediate public comment from Anthropic, it has been reported.

Reaction and next steps

Developers mostly celebrated a rare inside look at a sophisticated AI agent stack; security researchers warned of the potential for misuse. Will Anthropic patch the npm configuration, rotate secrets and rethink build‑pipeline hygiene? For now the global developer community has a sudden, detailed snapshot of an advanced agent architecture — courtesy of a single misconfigured source‑map.

AISmartphonesRobotics
View original source →