An LLM Wiki That Keeps Context When the Session Ends
This post describes how to collect the working context of AI agents such as Claude Code in a single Obsidian wiki and pull it back in from a new session. It adapts the LLM Wiki pattern Andrej Karpathy published for day-to-day work, and between 2026-08-20 and 2026-11-08 it accumulated 166 wiki pages and 140 log entries. The focus is on what the user sets up and what they say, with a short…
This article explains how to create an LLM Wiki using Obsidian for AI agents like Claude Code. The wiki keeps track of context across sessions, eliminating the need for the agent to be briefed from scratch each time. The setup involves creating a vault in Obsidian and running an install script. The agent uses the CLAUDE.md file in the wiki folder to understand what to do, based on phrases the user says.
The wiki is organized into three parts: raw/ for source material, wiki/ for decisions and project context, and conversations/ for ongoing session handoffs. Users can interact with the agent using five commands, such as "check the wiki" to summarize context and "save this to the wiki" to add new information. The agent only saves information that can be reused, contributes to the project, or could help resolve a failure.
It does not save one-off answers, unverified guesses, or intermediate thinking. The wiki is a git repository, so every save is committed and pushed, with the agent as the author. Linting the wiki checks for issues like copied source text or broken links. The system has been used to keep track of testing tool results, ensuring that verified information is properly documented.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
