obsidian-mcp-server and obsidian-http-mcp
These two projects are **competitors**, as they are both implementations of an MCP (Model Context Protocol) server for Obsidian vaults, offering similar functionality to expose vault data to AI agents and development tools.
About obsidian-mcp-server
cyanheads/obsidian-mcp-server
Obsidian Knowledge-Management MCP (Model Context Protocol) server that enables AI agents and development tools to interact with an Obsidian vault. It provides a comprehensive suite of tools for reading, writing, searching, and managing notes, tags, and frontmatter, acting as a bridge to the Obsidian Local REST API plugin.
Implements a modular TypeScript architecture leveraging stdio transport for MCP communication with automatic reconnection, backed by an intelligent in-memory vault cache that provides fallback search functionality if the REST API becomes unavailable. Supports flexible file targeting (by path, active note, or periodic notes) and atomic frontmatter operations, with structured logging, Zod validation, and optional HTTP transport via Hono for alternative deployment scenarios.
About obsidian-http-mcp
NasAndNora/obsidian-http-mcp
First HTTP-native MCP server for Obsidian - Compatible with any MCP client (Claude Code, Codex, Gemini, etc.) without stdio bugs
Implements HTTP transport bridging between MCP clients and Obsidian's Local REST API plugin, bypassing stdio reliability issues affecting Claude Code CLI. Features intelligent caching (70% fewer API calls), fuzzy search with typo tolerance, soft-delete protection, and sub-200ms response times even with 1000+ notes. Designed for cross-platform setups (Windows/WSL2/Linux) with straightforward 1-minute configuration using Node.js.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work