Anthropic accidentally shipped 512K lines of Claude Code's TypeScript source in a package update, exposing unreleased features including a Tamagotchi-style AI pet and always-on agent.
Anthropic's Claude Code 2.1.88 update accidentally included a source map file containing its full TypeScript codebase — over 512,000 lines of code. Users immediately posted the code publicly, and a GitHub repository containing the leak amassed 50,000+ forks before Anthropic patched the release. The leaked code revealed unreleased features including a Tamagotchi-style AI 'pet' and an always-on background agent. Anthropic confirmed no customer data or credentials were exposed, attributing the incident to human error in the release packaging process.
The leaked source reveals architectural decisions in Claude Code: how the agentic pipeline is structured, what guardrail logic looks like, and what features are in active development. The always-on agent pattern in the leak signals Anthropic is moving toward persistent background execution — which changes how you'd design integrations. The 50,000+ forks mean this code is now a de facto reference implementation for agentic coding tool architecture.
Browse the leaked GitHub repo's agent orchestration files this week to understand how Anthropic implements tool-use loops and guardrails — then compare against your own agentic pipeline to identify gaps or architectural shortcuts you can steal.
Search GitHub for 'claude code source leak' or navigate to the most-forked repository from the leak
Tags