Browse / Second brain / LLM wiki / obsidian-llm-wiki-assistant
obsidian-llm-wiki-assistant
HurricaHjz
Second yourself.
Open HurricaHjz/obsidian-llm-wiki-assistant → Checked against the four rules.
ai-agentai-researchclaudeknowledge-baseknowledge-managementllmobsidianpersonal-assistantpkmsecond-brain
Built for Claude Code
ships .claude CLAUDE.md and nothing for other agents.
In its own words
An AI that remembers you: an operating system for agents (Claude Code) on a self-maintaining, cited Obsidian wiki that is its long-term memory, after Karpathy's LLM-wiki pattern. Research assistant today; a shared brain for multi-agent work. Pre-1.0, in active development.
A personal assistant that remembers you, and a second brain that maintains itself.
Tony Stark never had to reintroduce himself to Jarvis. With today's AI you do it every day: close the chat, and everything you explained is gone.
The problem is memory, not intelligence. An agent does all its thinking in a context window that empties the moment a conversation ends. Think of the model as a CPU and the context window as its RAM, fast but volatile; with no disk underneath, everything you teach it is quietly thrown away. This project adds that disk, with a layered wiki as the bridge. You drop raw sources into your Obsidian vault, the agent compiles them into linked notes it looks after entirely on its own, and whenever a conversation needs something, the right notes flow back up into working memory. Everything stays plain Markdown, yours to open, read and correct. Feed it for a month and the assistant stops being a stranger; a year in, nothing else will know your work the way it does.
What you build on that memory is up to you:
The repository's own description, then 14% of its README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.