Anthropic's Claude Code Hides Internal Operations, Sparking Developer Backlash and Open-Source Solution
Key Takeaways
- ▸Anthropic modified Claude Code to hide file paths and internal reasoning by default, triggering developer backlash over reduced transparency
- ▸The company's suggested verbose mode workaround is viewed as inadequate, with no plans to restore original functionality
- ▸Open-source tool Claude-Devtools provides a solution by reading local session logs to reconstruct context windows, token usage, and tool calls
Summary
Anthropic has updated its Claude Code command-line tool to hide file paths and internal reasoning by default in an effort to reduce visual clutter and improve user-friendliness. The change has sparked significant backlash from software developers who argue that the lack of transparency compromises their ability to properly supervise the AI coding assistant's operations. While Anthropic suggests using verbose mode as a solution, many engineers find this setting either too noisy or insufficient for professional coding oversight, and the company has indicated no plans to revert to the original behavior.
In response to this controversy, an open-source tool called Claude-Devtools has emerged as a workaround. The free desktop application reconstructs and visualizes Claude Code's hidden background activities by reading raw session logs stored locally in the ~/.claude/ directory, without modifying Claude's behavior. Key features include context reconstruction with per-turn token attribution breakdowns, compaction visualization showing how the context window fills and compresses over time, a rich tool call inspector with syntax-highlighted code and inline diffs, and support for SSH remote sessions.
Claude-Devtools works across Linux, macOS, and Windows, with a platform-independent Docker version also available. The tool uses native file system watchers to provide near real-time updates as logs are written, making it useful for both active monitoring and post-mortem debugging. The controversy highlights the tension between simplifying AI tools for broader accessibility and maintaining the transparency that professional developers require for effective oversight and debugging of AI-assisted coding workflows.
- The tool offers real-time monitoring capabilities and works without modifying Claude Code's behavior, supporting multiple platforms
- The incident reveals ongoing tension between user-friendly AI interfaces and the transparency needs of professional developers
