What changed in the second release of 20 November 2026
20 November 2026 · Semurg · 1 min read · Markdown
What changed in the second release of 20 November 2026
Your agent picks up its session brief. When your agent stores what it learned with a session name, Semurg now builds a short brief for that session at the moment of the write. When the next session starts, the agent reads that brief in one lookup instead of searching its whole memory. The Claude Code plugin does this for you: its prompt hook asks for the brief first and adds it to the prompt only when it has changed, so an unchanged brief is not added to the prompt again.
Take your memory with you, history included. Export a memory, import it on another node, export it again, and you get back the same entities and facts, including every earlier value of an attribute and its date, in order. (Checked in this release's test suite.)
Still coming. Large open models served on your own servers: work continues, and a speed appears here only when a published release has measured it. Known issues are listed on the agents page.
What Semurg is: a memory that lasts for your AI agents, on a machine you run.
Why use it: your agent recalls the facts it needs, each with its source, instead of reading the project again.
What it solves today: agents that forget between sessions, and data that must stay on your own machines.
Try it: curl -fsSL https://one.semurg.io/install | sudo bash, first node free forever.
This post was written with AI assistance. Every result in it is checked against a published release or record before the post goes out.