“Claude Code,” developed by Anthropic, has rapidly become a focal point in the technology world, akin to the initial buzz around ChatGPT. Conceived by Boris Cherny, a former Meta engineer, during his initial weeks at Anthropic, Claude Code began as a side project and has evolved into a powerful tool widely utilized by over 80% of Anthropic’s engineers in their daily tasks. Its public release in May 2025 marked a significant milestone for the software.
Cherny, in various discussions and notably on the Pragmatic Engineer podcast and via an X post, shared insights into his personal setup of Claude Code at Anthropic. Despite describing his arrangement as relatively “vanilla,” the way Cherny employs Claude Code offers numerous optimizations that significantly enhance productivity and functionality beyond typical usage.
A unique aspect of Cherny’s usage of Claude Code is his management of multiple AI sessions simultaneously. Unlike the common practice of handling a single session at a time, Cherny runs up to 15 Claude Code sessions. He organizes five of these within his terminal, identified through numbered tabs, and an additional five to ten through the web interface. This setup not only maximizes his coding efficiency but is further streamlined through the use of iTerm2 system notifications, which alert him when sessions require input or are completed. Additionally, Cherny leverages mobile devices to initiate and monitor Claude Code sessions, enhancing flexibility and continuous workflow regardless of location.
Another significant feature of Cherny’s approach is the use of multiple Command-Line Interface (CLI) sessions. This methodology allows tasks to be delegated and monitored across different interfaces, which facilitates a seamless transition of tasks between local and web environments, and even allows for remote management through mobile devices.
Highlighting the adaptability and user-oriented design of Claude Code, Cherny emphasizes the importance of customization through the platform’s “Plan mode.” This feature allows for an in-depth analysis and planning phase where the code can be scrutinized and optimized before final execution, ensuring accuracy and minimizing the need for later corrections. This mode is crucial for maintaining control over the AI’s output, ensuring it aligns with the user’s specific requirements.
The integration capability of Claude Code is also notable. Through the use of MCP (Multi-Connection Protocol) servers, Claude Code can seamlessly interact with various tools and platforms that developers are already using, such as Slack, GitHub, and Google Drive. This integration not only simplifies the workflow but allows Claude Code to perform actions within these platforms, substantially reducing the need to switch contexts and thereby increasing efficiency. Cherny’s team, for instance, uses a shared .mcp.json file to ensure uniform integrations across all team members, which standardizes their environment and streamlines collaborative efforts.
Documentation and continuous learning are key components of Cherny’s strategy with Claude Code. Each project within Anthropic includes a CLAUDE.md file, where team members can document issues, behaviors, or specific code preferences necessary for Claude Code to operate effectively within their workflow. Over time, this file is updated, allowing the AI to learn from past errors and refine its operations, thereby progressively reducing repetitive mistakes and enhancing code consistency.
A crucial takeaway from Cherny’s setup is the versatility and adaptability of Claude Code, encouraging users to tailor the tool to their unique coding contexts and challenges. By running multiple parallel sessions, integrating with multiple tools, and maintaining rigorous, ongoing documentation, users can significantly amplify their productivity and code management efficiency.
Cherny’s innovative use of Claude Code demonstrates how the tool’s capabilities can be maximized beyond its basic functionality. His methods serve as a framework that, while adaptable to different scales of operation from individuals to large teams, provides valuable insights into managing complex coding environments more effectively. As AI continues to integrate more deeply into software development, understanding and implementing such enhanced workflows will become increasingly important, offering significant advantages in terms of efficiency, accuracy, and overall project management.
Read the full post on xda-developers.com
