Core Features
Designed for safety, performance, and developer experience. Every feature gives AI intelligent access without compromising your data.
Powerful Search
Fast full-text search with multi-word matching and BM25 relevance ranking. AI can locate notes by content, tags, or metadata instantly.
Safe Frontmatter Handling
YAML parser validates and preserves formatting, ensuring safe atomic updates to note metadata.
File Operations
Read, write, and manage notes safely. Create, update, and organize your vault with AI assistance.
Security First
Path traversal protection and safe file operations. Controlled AI access through MCP protocol.
Node.js Compatible
Built with Node.js for broad compatibility and ecosystem support.
Token Optimized
Minified JSON field names and compact responses. Less token usage means faster, cheaper API calls.
TypeScript
Fully typed for excellent developer experience.
Open Source
MIT licensed and community driven.
Complete Toolkit
14 MCP tools for vault management: read/write/patch/move files, search content, manage tags, update frontmatter, vault stats, and more. Built for AI assistant integration.
Multi-Platform
Works with Claude Desktop, ChatGPT+ Desktop, OpenCode, Gemini CLI, OpenAI Codex, Cursor, Windsurf, IntelliJ, and other MCP-compatible AI platforms.
Why Choose MCP-Vault?
See how MCP-Vault compares to plugin-based alternatives and direct file manipulation. Purpose-built for Obsidian means better safety, performance, and intelligence.
FAQ
Common setup and safety questions.
Does my data leave my computer?
Vault files stay on your machine. MCP-Vault reads and writes local files directly. Your AI provider only sees content that your client sends to it.
Does Obsidian need to be running?
No. MCP-Vault uses filesystem access, so it works even when Obsidian is closed.
Can I use multiple vaults?
Yes. Configure multiple MCP server entries, each pointing to a different vault path.
What file types are supported?
Read and write tools support .md, .markdown, .txt, .base, and .canvas files. Directory listing can include other filenames (like images or PDFs), but non-note files are not read as notes.
Is search semantic?
Search is lexical full-text search with multi-word matching and BM25 relevance ranking. It does not use embeddings or vector indexes.
What if the AI makes a mistake?
Use version control or vault backups for recovery. Deletions require path confirmation, and write operations are scoped to your selected vault.