Anthropic has confirmed the leak of more than 512,000 lines of source code for its Claude Code tool. The disclosure reveals disabled features hinting at future developments, including a persistent background agent called Kairos. Observers examining the code also found references to stealth modes and a virtual assistant named Buddy.
Anthropic confirmed on April 1 that its Claude Code source code had leaked, spanning over 512,000 lines across more than 2,000 files. The company builds scaffolding around its proprietary Claude model for coding tasks. As first reported by TechRadar, the breach spilled details on inactive features planned for future releases. Ars Technica detailed how experts sifted through the code to uncover these insights yesterday, April 1, 2026. Kairos emerges as a key planned feature: a persistent daemon operating in the background, even with the terminal closed. It uses periodic “tick” prompts and a “PROACTIVE” flag for “surfacing something the user hasn’t asked for and needs to see now.” The system relies on a file-based memory to track user preferences, collaboration styles, and work context across sessions. An AutoDream system would consolidate memories during idle periods, scanning transcripts for new information, pruning duplicates or outdated entries, and synthesizing durable memories to speed up future sessions. Separately, an “Undercover mode” aims to let AI contribute to public open-source repositories without revealing its nature, omitting phrases like “Claude Code” or AI attributions. Buddy, a Clippy-style watcher with 18 randomized species in ASCII art, was slated for a teaser launch between April 1 and 7, with a full rollout in May. Other referenced features include UltraPlan for Opus-level planning, Voice Mode for direct chats, Bridge mode for remote sessions, and a Coordinator for multi-worker tasks. The leak's impact on these timelines remains unclear.
