<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Codex on SPERIXLABS</title><link>https://sperixlabs.org/tags/codex/</link><description>Recent content in Codex on SPERIXLABS</description><generator>Hugo</generator><language>en-us</language><copyright>SPERIXLABS</copyright><lastBuildDate>Sat, 11 Apr 2026 18:10:00 +0000</lastBuildDate><atom:link href="https://sperixlabs.org/tags/codex/index.xml" rel="self" type="application/rss+xml"/><item><title>What Leaves Your Workstation When You Use an LLM Coding CLI</title><link>https://sperixlabs.org/post/2026/04/what-leaves-your-workstation-when-you-use-an-llm-coding-cli/</link><pubDate>Sat, 11 Apr 2026 18:10:00 +0000</pubDate><guid>https://sperixlabs.org/post/2026/04/what-leaves-your-workstation-when-you-use-an-llm-coding-cli/</guid><description>&lt;h2 id="tldr"&gt;TL;DR&lt;a class="anchor" href="#tldr" aria-label="Anchor to this section"&gt;#&lt;/a&gt;
&lt;/h2&gt;
&lt;p&gt;I put five LLM coding CLIs through a transparent TLS-intercepting proxy for about three and a half hours in a lab environment and captured every HTTP(S) transaction they made. The tools were &lt;strong&gt;Claude Code&lt;/strong&gt;, &lt;strong&gt;GitHub Copilot CLI&lt;/strong&gt;, &lt;strong&gt;Cursor CLI&lt;/strong&gt;, &lt;strong&gt;OpenAI Codex CLI&lt;/strong&gt;, and &lt;strong&gt;Opencode CLI&lt;/strong&gt;. The capture ran on an Apple Silicon MacBook (darwin/arm64) on 2026-04-11 and produced 302 transactions across 18 hosts.&lt;/p&gt;
&lt;p&gt;A few things stood out:&lt;/p&gt;</description></item></channel></rss>