AgriciDaniel/claude-obsidian

Self-organizing AI second brain for Obsidian + Claude Code. Drop any source and Claude reads, links, and files it into one connected knowledge graph of plain Markdown you own. AI note-taking, personal knowledge management (PKM), and an open-source Notion alternative. Based on Karpathy's LLM Wiki pattern.

What it solves

claude-obsidian transforms a standard Obsidian vault into a local-first, provenance-aware knowledge system. It solves the problem of AI-generated notes becoming disconnected from their original sources, preventing "hallucinations" by ensuring every claim is grounded in evidence and that the user maintains full ownership of their files in plain Markdown.

How it works

The system operates as a plugin for Claude Code (and other compatible Agent Skills hosts). It implements a repeatable loop of capturing local sources, grounding claims in source-cited pages, and connecting knowledge through linked notes and visual maps.

Key technical mechanisms include:

  • Transaction-safe writes: Every change to the vault is treated as a recoverable transaction with a pre-flight plan and a SHA-256 hash approval process to prevent silent overwrites.
  • Provenance tracking: It maintains source and claim ledgers to track authority, confidence, and contradictions.
  • Local-first architecture: The vault remains a directory of Markdown and JSON files, avoiding cloud databases or hidden caches.
  • Retrieval: It uses deterministic BM25 retrieval, with optional cosine reranking for more precise evidence gathering.

Who it’s for

Researchers, knowledge workers, and Obsidian users who want to leverage LLMs to synthesize information without losing the original source context or sacrificing data sovereignty.

Highlights

  • 15 Integrated Skills: Includes specialized tools for ingestion (wiki-ingest), querying (wiki-query), linting (wiki-lint), and automated research (autoresearch).
  • Methodology Support: Supports multiple filing conventions including Zettelkasten, PARA, and LYT.
  • Visual Mapping: Integrates with Obsidian Canvas for creating visual knowledge maps.
  • Strict Trust Boundary: Explicitly separates the product code from the user's knowledge vault to ensure data security.
  • Evidence-Based: Requires two independent sources for high-risk claims and prefers grounded refusals over invented citations.

Related

  • Project
    jaredrhod/ai-memory-vault

    A system that turns an Obsidian vault into a persistent, external memory for AI agents, allowing them to remember information across sessions without context limits.

  • Project
    smixs/agent-second-brain

    An always-on AI personal assistant that converts Telegram voice notes and messages into a structured, linked knowledge graph in a self-hosted Obsidian vault.

  • Project
    eugeniughelbur/obsidian-second-brain

    A persistent AI memory system for Obsidian that allows LLMs to read, write, and automatically maintain a self-updating knowledge vault across multiple AI platforms.

  • Project
    freestylefly/wesight-obsidian

    An Obsidian plugin that integrates local AI agents like Claude Code and Codex for vault-aware chat and inline editing, featuring a comprehensive workbench for publishing notes to social media and professional platforms.

  • Project
    RAIT-09/obsidian-agent-client

    An Obsidian plugin that integrates AI agents via the Agent Client Protocol, allowing agents to read and edit notes in your vault without copy-pasting.