claude-obsidian

I came across `AgriciDaniel/claude-obsidian` (GitHub Trending), an open-source “second brain” for Obsidian built around Claude Code.

What it does, in practical terms: instead of keeping notes and retrieval separate, it drops sources into Obsidian and lets Claude read them, then links and organizes the results into a connected knowledge graph stored as plain Markdown you own. The workflow is positioned as a Notion alternative for personal knowledge management (PKM), using the same general “LLM Wiki” pattern popularized elsewhere (Karpathy’s LLM Wiki idea).

Where I think it’s useful is when you already have a growing Obsidian vault (docs, papers, scratch notes) and you want the assistant to progressively curate it. A concrete engineer flow I’d use: after finishing a project phase, add raw outputs (meeting notes, design docs, logs) to the vault, then run Claude Code to map entities/ideas and produce consistent links back into the relevant Markdown files.

If you want to evaluate it quickly, here’s what to look at first:

- The repo’s setup/walkthrough to see how Claude Code is wired to your Obsidian vault - How it decides what becomes a node/link vs. what stays as a raw note - The entry point scripts/config that control “drop sources → read → link → write” - Example outputs so you can sanity-check graph density and formatting

Overall, this is the kind of tooling I like: small, workflow-driven, and geared toward owning the stored artifacts (Markdown) instead of hiding everything behind a hosted product.