AgriciDaniel/claude-obsidian

Turn scattered documents into a connected knowledge graph with AI

Drop papers, articles, or meeting notes into a single folder, and the AI reads the content and organizes it into interconnected notes. When questions arise later, it finds answers based on those notes, and you can save important insights back into the system. Eventually, all knowledge accumulates as plain text files on your computer, forming a personal knowledge base you can open at any time.

How materials accumulate

This tool does more than store text; it builds relationships between materials. When you place source materials in the inbox/ folder, the AI reads the content and creates linked notes with explicit sources. The original files remain preserved, so you can always trace where a conclusion came from. As new ideas arrive, they connect to existing notes and expand like a graph, keeping information organized in a single system rather than scattered.

15 features for handling knowledge

Users can perform various tasks by conversing in natural language or using specific commands.

• Collection and organization wiki-ingest reads materials in inbox/ and converts them into linked pages. wiki sets up a new repository or adopts an existing Obsidian vault.
• Querying and retrieval wiki-query answers questions based on information in the repository without modifying files. wiki-retrieve performs local BM25 search to find relevant notes.
• Saving and management save stores good answers or insights as notes. wiki-lint identifies health issues such as broken links, isolated pages, or empty sections.
• Research and visualization autoresearch performs web research with user consent. canvas creates or updates Obsidian Canvas boards to arrange knowledge visually.

Design for data safety

The tool handles all changes carefully to protect user notes. Before the AI modifies a note, it records the file's fingerprint (SHA-256 hash) and applies the change plan as a single transaction. If the user edits a note simultaneously in Obsidian, the fingerprints will not match, causing the AI's record to stop as a conflict rather than overwriting the user's edits. Additionally, if power is suddenly cut, the journal allows rolling back interrupted changes, reducing the risk of data loss.

Installation and requirements

Using this tool requires Python 3.11 or higher and the latest version of Claude Code. Installing Obsidian is necessary for visual verification but is not mandatory. The installation process is completed with two commands, and you can adopt an existing Obsidian vault as is. In Windows environments, using WSL is recommended for file writing tasks.

Who this tool is for

It is suitable for researchers or students who read many materials and want each note to point to its source. It is also good for writers or creators who want ideas from videos, articles, and books to be connected rather than scattered in folders. It is useful for developers or teams who manage project knowledge in Markdown and want the AI to search, cite, and organize it. However, this tool is not an automatic transcription recorder or a cloud synchronization service, and additional tools are needed for full text extraction from PDFs or web pages, so users should be aware of these limitations.