Inside Claude Code’s leaked source: swarms, daemons, and 44 features Anthropic kept behind flags
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A 59.8MB source map accidentally shipped with Claude Code v2.1.88 exposed 512,000 lines of unobfuscated TypeScript across 1,900 files before Anthropic pulled the package. Analysis of the leaked code reveals Claude Code is architected as an agent operating system with a 40+ tool syscall-style permission layer, multi-agent swarm orchestration (gated behind feature flags), and two unreleased background services: KAIROS (an autonomous daemon that proactively acts on a schedule) and autoDream (a memory consolidation sub-agent). The leak also exposed 44 compile-time feature flags hiding finished but unreleased capabilities including voice mode, ULTRAPLAN, and a Tamagotchi companion pet. Internal model codenames were revealed, including Capybara (Claude 4.6/Mythos), Fennec (Opus 4.6), and an unreleased model called Numbat. The incident coincided with a supply-chain attack on the Axios npm package, highlighting that build pipelines are now a critical attack surface for AI tooling teams.
Table of contents
The tool system as a syscall interfaceMulti-agent swarms and process orchestrationKAIROS and the dream system as background servicesCompile-time feature gating as a shipping strategyWhat’s nextSort: