<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Oxus on Calogero Sanfilippo</title><link>https://calogerosanfilippo.it/tags/oxus/</link><description>Recent content in Oxus on Calogero Sanfilippo</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Tue, 03 Nov 2026 16:20:14 +0000</lastBuildDate><atom:link href="https://calogerosanfilippo.it/tags/oxus/index.xml" rel="self" type="application/rss+xml"/><item><title>oxus: Spotlight search for MCP clients, built test-first with Claude</title><link>https://calogerosanfilippo.it/posts/oxus-released/</link><pubDate>Tue, 03 Nov 2026 16:40:00 +0100</pubDate><guid>https://calogerosanfilippo.it/posts/oxus-released/</guid><description>&lt;h2 id="why"&gt;&#10; Why?&#10; &lt;a class="heading-link" href="#why"&gt;&#10; &lt;i class="fa-solid fa-link" aria-hidden="true" title="Link to heading"&gt;&lt;/i&gt;&#10; &lt;span class="sr-only"&gt;Link to heading&lt;/span&gt;&#10; &lt;/a&gt;&#10;&lt;/h2&gt;&#10;&lt;p&gt;I wanted to find out what test-driven development feels like when the other half of the pair is an AI. I didn&amp;rsquo;t want a code generator: I wanted a pair in the XP sense, someone to shape the design with, keep the test list honest and question each step.&lt;/p&gt;&#10;&lt;p&gt;An experiment like that needs a problem small enough to finish and real enough to be worth finishing. I chose an MCP server for Spotlight. I had been meaning to learn the &lt;a href="https://modelcontextprotocol.io" class="external-link" target="_blank" rel="noopener"&gt;Model Context Protocol&lt;/a&gt;, and the problem was right there: every Mac already keeps an index of its files, yet an agent looking for a document usually crawls the disk with &lt;code&gt;find&lt;/code&gt;.&lt;/p&gt;</description></item></channel></rss>