<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: augustus henry</title>
    <description>The latest articles on DEV Community by augustus henry (@augustushenry2007).</description>
    <link>https://dev.to/augustushenry2007</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3866320%2F9908192d-d1c2-416e-800e-6ead29924ed6.jpg</url>
      <title>DEV Community: augustus henry</title>
      <link>https://dev.to/augustushenry2007</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/augustushenry2007"/>
    <language>en</language>
    <item>
      <title>I kept getting wrecked by Claude API bills. So I built a middleware layer.</title>
      <dc:creator>augustus henry</dc:creator>
      <pubDate>Tue, 07 Apr 2026 17:45:13 +0000</pubDate>
      <link>https://dev.to/augustushenry2007/i-kept-getting-wrecked-by-claude-api-bills-so-i-built-a-middleware-layer-4mbe</link>
      <guid>https://dev.to/augustushenry2007/i-kept-getting-wrecked-by-claude-api-bills-so-i-built-a-middleware-layer-4mbe</guid>
      <description>&lt;p&gt;I'm a music composer. I run a sonic branding studio called Fable Audio. I am not, by any definition, an infrastructure engineer.&lt;br&gt;
And yet here I am building API tooling. Because Claude kept billing me into mild panic.&lt;/p&gt;

&lt;p&gt;The problem&lt;br&gt;
I was building Claude-powered workflows and token usage kept compounding faster than I expected. Each call in an agentic chain drags context from the last one. By call four or five you're paying for a lot of history the model doesn't need.&lt;br&gt;
Every fix I found required rewriting app logic. I wanted something that handled it at the layer between my app and the API — without touching my code.&lt;/p&gt;

&lt;p&gt;What I built&lt;br&gt;
Tokenly is an optimization layer that sits between your app and the Anthropic API. BYOK — your key goes straight to Anthropic. Still early, still figuring it out.&lt;br&gt;
tokenly.onrender.com&lt;/p&gt;

&lt;p&gt;If you're hitting the same wall on Claude API costs, especially on multi-turn or agentic workflows, I'd genuinely appreciate feedback on whether this is useful to you.&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>claudeai</category>
      <category>llm</category>
    </item>
  </channel>
</rss>
