How to Use Claude Code With Your Obsidian Notes

How do you connect Claude Code to Obsidian?

Use the Obsidian Second Brain plugin to give Claude Code vault-specific instructions and a connector to your notes. Start by installing the plugin, configuring the local Obsidian connection and asking a read-only question. Then add capture, research and organization workflows once you have checked the returned sources.

The plugin supplies conventions for wikilinks, frontmatter, callouts, templates and Maps of Content. Claude may already understand some Obsidian syntax, but a shared procedure makes your vault's expectations explicit. Check edits against those conventions; advisory hints are not a file-format validator or a backup.

This is one plugin in a broader marketplace of Claude Code plugins : see Rhize plugins for Claude Code for the full catalog and installation instructions.

What problem does it solve?

A vault full of notes is not the same thing as a working second brain. Most Obsidian vaults accumulate captured content : clipped articles, meeting notes, half-finished ideas : without the connective tissue that makes a knowledge system useful: links between related notes, consistent tagging, MOCs that summarize a topic, and a repeatable process for turning fleeting notes into permanent ones.

This plugin addresses that gap on three fronts. First, it teaches Claude actual PKM methodologies (Zettelkasten, PARA, progressive summarization) so it applies them instead of writing generic bullet points into a markdown file. Second, it wires Claude into your vault through the Obsidian MCP server and CLI so it can read, write, search, and reorganize notes directly rather than guessing at file paths. Third, it adds semantic search via qmd, so Claude can find conceptually related notes even when they share no exact keywords : a real limitation of plain full-text search once a vault grows past a few hundred notes.

What skills does the plugin ship?

The current package combines knowledge-workflow skills with format references and a knowledge compiler. Rather than memorizing the inventory, choose the task: recall an earlier decision, capture a source, inspect vault health or preview a compiled knowledge page. Skills may load when relevant, and explicit commands are useful when you want a particular workflow.

Format skills provide instructions for files such as Obsidian Bases. For example, ask for a project tracker with title, status and due-date columns, then confirm that the host discovered or explicitly invoked obsidian-bases. Inspect the resulting .base file and open it in Obsidian to validate its structure and behavior; skill discovery and correct output are not guaranteed.

What commands does it provide?

The current command set covers research, recall, capture, daily notes, search, health checks, setup and compiled knowledge. The table below lists the entry points. A command can combine several tools, so check its prerequisites before assuming the whole workflow is available.

How does the research pipeline work?

/vault-research is the command that best shows what the plugin is for. Point it at a URL and it clips the content with Defuddle, creates a properly formatted note with metadata, writes a summary, searches your vault for related notes, suggests connections, and logs the research in your daily note : one command, one pipeline, several skills working together.

Point it at a topic instead of a URL and the behavior changes: it searches the web, lets you pick sources, processes each one through the same pipeline, and offers to build a synthesis note once it's done. In our own vault at Rhize Media, we use this for technical deep-dives : running /vault-research transformer architecture advances produces a properly linked note set instead of a wall of text we'd have to break apart and connect manually.

Three example prompts that show the range of the command set:

• /vault-research https://example.com/interesting-article : clip, summarize, and connect a single source
• /vault-connect recent : find your most isolated recent notes and weave them into the graph
• /vault-recall What decisions did we make about the API architecture? : ask the vault a question and get a cited, synthesized answer instead of a list of search hits

/vault-recall searches for relevant notes, reads the matches and composes an answer with wikilink citations. qmd can add hybrid retrieval when configured; otherwise the workflow uses available keyword search. Check the cited passages and ask the assistant to identify uncertainty or conflicting notes. A semantic match is not proof that a statement is supported.

How does vault health checking work?

/vault-align runs a structured audit against five dimensions: structure (folder organization), connectivity (links and MOCs), consistency (frontmatter and tags), processing (whether captures actually become permanent notes), and plugin utilization. Running /vault-align check on a vault that's accumulated six months of unmanaged notes typically surfaces broken links, orphan notes with no incoming or outgoing connections, and tag drift : the same three problems, in our experience, that quietly make a vault stop being useful.

The command has four modes. check produces a scored report with prioritized suggestions. fix executes the single highest-impact improvement and re-checks the affected area. migrate builds and executes a full batch reorganization plan : file moves, tag consolidation, frontmatter standardization : with per-phase approval, using obsidian move so wikilinks don't break. plugins audits your installed community plugins against what your vault archetype needs.

What sets up the vault correctly the first time?

New vaults, or existing vaults you want to reorganize, start with /vault-setup. It interviews you about your role and workflow, audits your existing notes if you have any, generates a personalized folder structure, templates, dashboard, and MOC scaffold, and : for existing vaults : offers an opt-in migration to move everything into the new structure without losing links. It also writes a _vault-setup-log.md note that /vault-align later reads to know what your vault is supposed to look like.

What are the setup prerequisites?

Set up the connection before asking for edits. Obsidian must be running for the bundled Local REST API connector. The CLI supports direct vault operations; Defuddle is needed for clipping and qmd is optional for semantic search. Missing dependencies should be reported for the affected capability, not treated as a successful run.

• 1. Enable Obsidian's Local REST API community plugin and obtain its API key privately. The bundled connector uses https://127.0.0.1:27124 without a trailing slash. Configure OBSIDIAN_API_KEY through the plugin's documented secret launcher; on macOS it can read the Keychain item claude-code:OBSIDIAN_API_KEY into the server process. Do not paste the value into a note or chat.
• 2. Register the Obsidian CLI under Settings > General > Command line interface, restart the terminal and check obsidian --version. The plugin documents v1.12.4 or later. Prefer its link-aware operations when moving notes, and verify links after any approved reorganization.
• 3. Add Defuddle only if you need web clipping: npm install -g defuddle. Start with one public source and inspect the captured note before processing a batch.
• 4. Add qmd only if semantic search is useful. Follow the current qmd setup in the plugin README and enable the qmd@qmd plugin. Index an explicitly approved note root, not the whole vault by default. Keep compiled-knowledge output and its private state outside qmd collections because the current compiler does not have an ACL-aware qmd export adapter.

First install the package in Claude Code: /plugin marketplace add https://github.com/Rhize-Media/rhize-plugins, then /plugin install obsidian-second-brain@rhize-plugins. Start a fresh session after installation or an update. Confirm that the connector is available, then ask: "Find the note about our last project decision and quote the supporting passage. Do not edit anything." A successful read with the expected source is the first checkpoint.

What happens automatically in the background?

The package supplies advisory hints around reading and writing vault Markdown. Write hints describe conventions such as wikilinks and frontmatter; read hints suggest related-note exploration. They are scoped to recognized vault paths and are not a security boundary. The old unconditional SessionStart banner was removed; relevant disclosure is handled by Context Manager when that separate plugin is installed. Verify host hook support and actual behavior after updates.

Who is this plugin for?

It's built for anyone running Obsidian as an active knowledge system rather than a folder of loose notes : researchers processing sources into permanent notes, writers building a MOC-linked idea graph, consultants tracking client history across engagements, or anyone doing periodic reviews of their own thinking. It assumes you already use Obsidian; it won't migrate you off another note tool, and the CLI- and MCP-dependent commands need Obsidian running in the background. If your vault is a flat folder of unlinked notes today, /vault-setup is the intended starting point, not /vault-research or /vault-align.

Getting started

After the read-only checkpoint, use /vault-setup to propose a structure or /vault-capture for a single approved note. Inspect the exact changes before a batch migration. Use /vault-align check to review the vault without reorganizing it, then approve a specific fix. For synthesized knowledge, /vault-compile preview creates a reviewable proposal; apply only the exact approved preview, keeping canonical sources separate from derived pages.

For the rest of the marketplace this plugin ships alongside, see Rhize plugins for Claude Code. The plugin source and installation instructions are on GitHub, and background on the underlying methodology is documented on Obsidian's own help site.

Sources and implementation references

Setup and capability reference: https://github.com/Rhize-Media/rhize-plugins/blob/main/obsidian-second-brain/README.md