Obsidian Local REST API is an MCP server that connects AI assistants to local Obsidian vaults via a local REST API endpoint. Designed for knowledge workers, researchers, and developers who rely on Obsidian for personal note-taking, the server allows Large Language Models to interact directly with markdown files, frontmatter metadata, and directory structures. Instead of relying on raw CRUD operations that quickly exhaust context windows, this server exposes high-level, AI-native tools. Users can perform paginated directory exploration, unified note writing with append, prepend, or overwrite modes, and automated upserting of notes with frontmatter. It also provides specialized routines for retrieving daily notes by relative dates, fetching recently updated documents, discovering related files through shared tags or wikilinks, and executing multi-scope vault searches. The server communicates locally over standard input/output (stdio), ensuring that knowledge bases remain private while enabling conversational agents to search, read, and organize personal documentation.
Category: Files, Documents & PDFs
Tags: knowledge-management, markdown, obsidian, rest-api
claude_desktop_config.json. 3. Add obsidian-vault under mcpServers with command npx and args ["obsidian-local-rest-api-mcp"]. 4. Under the env section, set OBSIDIAN_API_URL to your local API address (such as http://obsidian-local-rest-api.test or http://localhost:8000) and provide OBSIDIAN_API_KEY if authentication is configured. 5. Save the configuration and restart the client.Part of MCP Servers
You can run it directly using npx by specifying obsidian-local-rest-api-mcp in your client configuration. Alternatively, clone the repository from GitHub, run bun install to set up dependencies, compile the TypeScript source with bun run build, and run the server using node build/index.js.
You need Node.js 18 or later or the Bun runtime installed on your machine. You also need Obsidian running locally alongside the Obsidian Local REST API plugin accessible via HTTP on your configured port or local test domain.
The server allows AI models to browse vault directories with pagination, read and write files in overwrite or append modes, upsert notes with frontmatter, query daily notes, retrieve recently edited files, discover related notes through links, and search vault content.
Any client supporting the Model Context Protocol over standard input and output (stdio) can use this server. It includes documented configurations for Claude Desktop, VS Code with an MCP extension, and Cherry Studio.
Yes, this MCP server is an open-source project distributed under the MIT license on GitHub.