The article analyzes the data sent by popular coding agents and their implications for user privacy and data security. It focuses on seven coding agents, including Junie (a PyCharm plugin), Gemini CLI, Codex CLI, and others, evaluating their behavior through network traffic analysis. The research, conducted by Lucas Pye with guidance from senior engineers, aimed to understand how these tools operate in real-world scenarios, particularly concerning telemetry and potential risks like system prompt leakage. Using a transparent proxy setup, the study captured and decrypted traffic from these coding agents during various tasks, such as developing a leaderboard feature for a Flappy Bird game. The methodology involved setting up isolated environments to ensure that only the coding agents were active. Each agent was tested under different conditions: with telemetry opted out, opted in, and with telemetry domains blocked. The findings revealed that Junie and other agents processed significant data while interacting with external services, sometimes revealing sensitive information like AWS credentials. Key observations included that even when users opted out of telemetry, certain data still flowed through the agents, often including detailed context from user interactions. For Junie, the analysis showed that prompts were sent to an API rather than directly to the AI model, raising questions about data handling. Overall, the research highlights the need for vigilance when integrating AI tools into coding workflows, particularly regarding what data is being shared and how it might impact security and privacy.