What Claude Code Sends to the Cloud (rastrigin.systems)

🤖 AI Summary
A developer has created a proxy to investigate the data Claude Code sends to the cloud, revealing that it transmits more information than many users might expect, including extensive code, conversation history, and a significant system prompt. By intercepting traffic, the developer found that every response carries a request containing not just the user's messages but also the entire session's context, a massive system prompt of about 15,000-25,000 tokens, as well as definitions of tools and environmental details. This means that each user interaction incurs substantial bandwidth and token costs, as the entire conversation history is sent with every prompt. The significance of these findings lies in their implications for user privacy and data management. With each interaction, Claude Code shares substantial amounts of personal and project-related information with Anthropic's servers, raising concerns about confidentiality, especially for proprietary projects. Notably, the model employs Server-Sent Events (SSE) for real-time updates instead of WebSockets, simplifying its architecture. Developers should be mindful of the data they permit Claude Code to access, particularly when working with sensitive or confidential projects. These insights underline the trade-offs between convenience and privacy that come with using advanced AI coding tools.
Loading comments...
loading comments...