<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title>Memory - Tag - EONYX Blog</title><link>https://eonyx.tech/blog/tags/memory/</link><description>Memory - Tag - EONYX Blog</description><generator>Hugo -- gohugo.io</generator><language>en</language><managingEditor>eonyx.tech@proton.me (EONYX)</managingEditor><webMaster>eonyx.tech@proton.me (EONYX)</webMaster><copyright>This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.</copyright><lastBuildDate>Wed, 17 Jun 2026 00:00:00 +0300</lastBuildDate><atom:link href="https://eonyx.tech/blog/tags/memory/" rel="self" type="application/rss+xml"/><item><title>EONYX memory: AGENTS.md, sessions and context compaction</title><link>https://eonyx.tech/blog/eonyx-memory/</link><pubDate>Wed, 17 Jun 2026 00:00:00 +0300</pubDate><author><name>EONYX</name></author><guid>https://eonyx.tech/blog/eonyx-memory/</guid><description><![CDATA[<div class="featured-image">
                <img src="/blog/eonyx-memory/cover.svg" referrerpolicy="no-referrer">
            </div><p>Every agent turn starts with the model receiving context. In EONYX it is built from three mechanisms: <strong>AGENTS.md</strong> — the project&rsquo;s persistent instructions, <strong>sessions</strong> — the full conversation history on disk, and <strong>auto-summarization</strong> — compaction of older turns once the history approaches the limit of the context window.</p>]]></description></item></channel></rss>