Hacker NewsWednesday · August 12, 2026FREE

What I learned by putting GitHub Copilot behind a MitM proxy

In a hands-on investigation, a developer routed GitHub Copilot through a man-in-the-middle proxy to capture and analyze its network traffic. The setup allowed them to see exactly what data the AI assistant transmits and receives during a typical coding session. The findings shed light on how Copilot builds its context—what snippets of code, comments, and file contents are included in each request—and how it manages memory across interactions. The proxy logs revealed the structure of the API calls, including headers and payloads, and showed how Copilot's client communicates with GitHub's servers. This technical deep dive provides a rare, behind-the-scenes view of the tool's inner workings, which could help developers understand its data usage and privacy implications. The article is based on the author's direct experience and includes specific observations from the captured traffic, though the full details are embedded in the source text.