obsidian-mcp
Enables MCP-compatible AI hosts to read, search, link, and write notes in a local Obsidian vault with sandboxed file access.
README
obsidian-mcp
A Model Context Protocol server that lets any MCP-compatible AI host (Cursor, Claude Desktop, Zed, etc.) read, search, link, and write notes inside a local Markdown / Obsidian vault.
What is this?
obsidian-mcp is an MCP server. MCP is an open
standard from Anthropic for connecting AI assistants to external tools and data
sources — think "USB-C for AI applications". Once this server is registered with
an MCP host you can ask the model questions like:
"What did I learn about MCP this week, and which of my notes link to it?"
…and the model will call this server's tools (search_notes, find_backlinks,
get_recent_notes, …) to answer using your actual notes.
Status
| Phase | Scope | State |
|---|---|---|
| 0 | Repo skeleton, CI, sample vault, server stub with get_note |
shipped |
| 1 | Sandboxed pathing, parser, reader, tests | shipped |
| 2 | Search, listings, backlinks, tag index | planned |
| 3 | Resources (notes as obsidian:// URIs) |
planned |
| 4 | Write tools (create_note, append_to_note) + atomic writes |
planned |
| 5 | Prompts (/weekly-review, /daily-note-template) |
planned |
Architecture
┌────────────────┐ stdio JSON-RPC ┌────────────────────────┐
│ MCP host │ ───────────────────► │ obsidian-mcp server │
│ (Cursor / │ │ (this repo) │
│ Claude / │ │ │
│ Zed) │ │ tools / resources / │
└────────────────┘ │ prompts │
└───────────┬────────────┘
│
sandboxed ▼
┌────────────────────────┐
│ Local Markdown vault │
└────────────────────────┘
Every caller-supplied path is funnelled through a single sandboxing function
(utils/pathing.py::safe_resolve) before any filesystem access, so the server
cannot be coerced into reading files outside the configured vault root.
Quickstart
1. Install
git clone https://github.com/darrenlopez/obsidian-mcp.git
cd obsidian-mcp
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
2. Try it against the bundled sample vault
OBSIDIAN_MCP_VAULT_PATH="$(pwd)/sample-vault" \
npx @modelcontextprotocol/inspector \
python -m obsidian_mcp
This launches Anthropic's official MCP Inspector
pointed at this server, so you can interactively call get_note and inspect
schemas without leaving your browser.
3. Register with Cursor
Add the following to your Cursor MCP config
(~/.cursor/mcp.json or the Cursor settings UI):
{
"mcpServers": {
"obsidian": {
"command": "python",
"args": ["-m", "obsidian_mcp"],
"env": {
"OBSIDIAN_MCP_VAULT_PATH": "/absolute/path/to/your/vault",
"OBSIDIAN_MCP_READ_ONLY": "false"
}
}
}
}
4. Register with Claude Desktop
Add the same block to ~/Library/Application Support/Claude/claude_desktop_config.json
on macOS (or the platform equivalent).
Configuration
All configuration is via environment variables (prefix OBSIDIAN_MCP_).
| Variable | Default | Purpose |
|---|---|---|
OBSIDIAN_MCP_VAULT_PATH |
(required) | Absolute path to the vault root. |
OBSIDIAN_MCP_READ_ONLY |
false |
When true, write tools are not registered. |
OBSIDIAN_MCP_MAX_FILE_KB |
1024 |
Max note size (KiB) returned by read operations. |
OBSIDIAN_MCP_INCLUDE_HIDDEN |
false |
Include dotfiles and .obsidian/ in listings/search. |
Tools (Phase 0 / 1)
| Tool | Description |
|---|---|
get_note(path) |
Read a single note, returning parsed frontmatter, tags, and outgoing wikilinks. |
The Phase 2+ tool surface (search_notes, list_notes, find_backlinks,
list_tags, get_recent_notes, …) is documented in the
architecture plan and tracked in STATUS.
Security
This server reads (and, in non-read-only mode, writes) files on your machine. Some choices that limit blast radius:
- Sandboxed paths. Every path is resolved through
safe_resolve(vault_root, user_input), which rejects absolute paths,..segments, and symlink escapes before any filesystem touch. - Read-only mode. Set
OBSIDIAN_MCP_READ_ONLY=trueand write tools are not registered at all. - Size limits.
OBSIDIAN_MCP_MAX_FILE_KBcaps the bytes returned by read operations to prevent DoS via huge files. - Hidden-file exclusion.
.obsidian/and dotfiles are skipped by default, so plugin secrets do not leak into model context. - No network. The server makes no outbound network requests of its own.
- No
shell=True, noeval. Anywhere.
Development
pip install -e ".[dev]"
ruff check .
ruff format --check .
mypy
pytest
The test suite includes adversarial path-traversal tests
(tests/test_pathing.py) — keep them green.
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.