ClaudeObsidian
Keeping a personal knowledge base current normally competes with doing the work the knowledge is for: every article read, every meeting, every decision has to be manually filed and cross-referenced or it decays into an unsearchable pile. claude-obsidian is a tool that hands that upkeep to an AI agent. It runs on top of Claude Code and an Obsidian vault: a source is dropped in, and the agent reads it, writes it into the appropriate note, and links it to everything related, so the vault organizes itself into a connected body of plain Markdown the user owns. Later questions are answered from everything the vault has ever absorbed. The project describes the effect as knowledge that compounds, and positions itself as an open-source alternative to Notion. It is an MIT-licensed project with roughly 9,700 stars built on Andrej Karpathy's LLM Wiki pattern.
Assess. The tool is mature and heavily adopted for a single-author project (around 9,700 stars, MIT-licensed, actively maintained into 2026), ships a substantial feature set (multiple methodology modes such as PARA and Zettelkasten, multi-writer file locking, and a set of Claude Code skills), and has a low barrier to trial. None of that is production evidence in the user's own workflow, so it stays in Assess. Inner position reflects the maturity and adoption rather than any confirmed first-person use; placement moves only after the user runs it against a real vault.
- Ingests an arbitrary source and files it into a connected Markdown vault, adding links to related notes automatically.
- Answers questions by drawing on everything previously absorbed, so the vault behaves as an accumulating knowledge base rather than a folder of clippings.
- Ships methodology modes (LYT, PARA, Zettelkasten, or generic) so the vault structure follows a chosen note-taking discipline.
- Adds per-file advisory locking so multiple agents can write to the same vault without corrupting notes.
The tool is the most-adopted implementation of the LLM-driven personal wiki idea, which makes it a useful reference point for anyone building or running the same pattern. It shares direct lineage with the LLM Wiki pattern that this radar's own wiki workflow descends from, so it illustrates where that approach leads when pushed to a full product: retrieval tuning, methodology presets, and concurrency safety layered on top of plain Markdown.
