Claude + Obsidian knowledge companion. This project provides an AI-powered knowledge engine that autonomously builds, organizes, and maintains a persistent, compounding wiki vault within Obsidian, eliminating manual filing. It targets users seeking an evolving, self-sufficient knowledge base, enabling knowledge to compound like interest.

How It Works

The system leverages LLMs (primarily Claude) to ingest user-provided sources, extracting entities, concepts, and updating cross-references before filing them into a structured Obsidian vault. Queries are answered by scanning the indexed knowledge base, with answers citing specific wiki pages rather than general training data. Automated vault maintenance (linting) identifies and flags issues like orphans or dead links, while a persistent "hot cache" maintains session memory across conversations. This approach positions it as a proactive knowledge engine rather than a reactive chat interface.

Quick Start & Requirements

• Primary install / run command: Recommended: git clone https://github.com/AgriciDaniel/claude-obsidian, cd claude-obsidian, bash bin/setup-vault.sh. Open the resulting folder in Obsidian and Claude Code. An alternative plugin installation method is also provided.
• Non-default prerequisites and dependencies: Obsidian, Claude Code (or compatible LLM interface). The setup-vault.sh script configures Obsidian's graph view, color scheme, and structure. Optional MCP (Multi-Claude Protocol) setup requires the Obsidian Local REST API plugin.
• Links: YouTube Demo, Deep dive blog post, AI Marketing Hub community.

Highlighted Details

• Autonomous creation, organization, and evolution of an Obsidian wiki vault.
• Persistent, compounding knowledge base built on the LLM Wiki pattern.
• Automated vault health checks (linting for orphans, dead links, gaps).
• Session memory maintained via a persistent "hot cache" (hot.md).
• Querying capabilities that cite specific wiki pages.
• Multi-agent support for parallel batch ingestion of multiple sources.
• Companion claude-canvas plugin for visual knowledge graphs, presentations, and flowcharts.
• Configurable autonomous research loop (/autoresearch).
• Supports six distinct "Wiki Modes" (e.g., Website, GitHub, Business, Research).

Maintenance & Community

Developed by Agrici Daniel. The project is supported by an active community, including the AI Marketing Hub (2,800+ members) and numerous YouTube tutorials. Related open-source tools like claude-seo and claude-ads are also available.

Licensing & Compatibility

The project is licensed under the MIT License, permitting broad use and modification. It is designed for Obsidian and integrates with Claude, with mentions of compatibility with Gemini, Codex, Cursor, and Windsurf.

Limitations & Caveats

While supporting multiple LLMs, the project is primarily optimized for Claude. Achieving full functionality, particularly for direct vault interaction via MCP, requires specific Obsidian plugin installations and configurations.