<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Grafana on El PoshoX</title><link>http://elposhox.dev/en/tags/grafana/</link><description>Recent content in Grafana on El PoshoX</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>© 2026 Az García Zúñiga</copyright><lastBuildDate>Fri, 05 Jun 2026 10:00:00 +0000</lastBuildDate><atom:link href="http://elposhox.dev/en/tags/grafana/index.xml" rel="self" type="application/rss+xml"/><item><title>What Each Claude Code Session Really Costs: Monitoring with OpenTelemetry</title><link>http://elposhox.dev/en/posts/claude-code-otel-homelab/</link><pubDate>Fri, 05 Jun 2026 10:00:00 +0000</pubDate><guid>http://elposhox.dev/en/posts/claude-code-otel-homelab/</guid><description>&lt;div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"&gt;
 From Claude Code&amp;rsquo;s env vars to a 34-panel Grafana dashboard. All running on a homelab with Talos Linux, ArgoCD, Mimir and Tempo.
&lt;/div&gt;

&lt;p&gt;I wanted to know how much each Claude Code session cost me, which models burned the most tokens, and how long I spent blocked waiting for tool call approvals. The answer was in the &lt;code&gt;OTEL_EXPORTER_*&lt;/code&gt; env vars that Claude Code recently started supporting. What wasn&amp;rsquo;t documented anywhere were the 5 problems I had to solve before any data actually showed up in my observability stack.&lt;/p&gt;</description></item></channel></rss>