obsidian-mcp-guard
Provides read/write file tools with lint validation to prevent agents from writing malformed Obsidian markdown, and constrains writes to a single configurable vault for security.
README
obsidian-mcp-guard
An MCP server for agent-safe Obsidian vault access. Provides read/write file tools with lint validation to prevent agents from writing malformed Obsidian markdown.
Why this exists
Most Obsidian MCP servers give agents direct write access with no markdown validation. Those that route through the Obsidian REST API gain some input sanitisation, but none validate content against Obsidian's markdown rendering rules before writing. obsidian-mcp-guard fills this gap: all writes are validated against Obsidian's markdown rules before they touch the vault, and if the content would render incorrectly, the write is rejected with a structured error explaining exactly which rule was violated. Writes can also be constrained to a single configurable vault path, giving agents a designated space to create and manage content on behalf of the user while preventing accidental or runaway writes to other vaults on the same filesystem. Directory traversal attacks are blocked at the path resolution layer, so a misconfigured, misbehaving, or prompt-injected agent cannot escape the write vault by constructing paths like Claude/../OtherVault/note.md
Features
- Read/list/create/update/delete/move notes via
HOST_VAULT_PATHon the host filesystem - Lint validation on all writes using mdlint-obsidian — blocks writes that violate Obsidian markdown rules (unclosed wikilinks, raw HTML, standard-markdown links, etc.)
- Write-vault isolation — writes are constrained to a single configurable vault; directory-traversal attacks are blocked on both read and write paths
- Composable —
create_vault_server()returns aFastMCPinstance that can be mounted into a larger server viaimport_server() - Pre-validation tool —
lint_notelets agents check content before committing a write
Installation
pip install obsidian-mcp-guard
For Claude Desktop users who don't want a manual install, uvx runs it directly with no setup:
uvx obsidian-mcp-guard
For local development:
python -m venv .venv
source .venv/bin/activate
pip install -e .
Configuration
| Environment variable | Default | Description |
|---|---|---|
HOST_VAULT_PATH |
(required) | Absolute path to the directory containing your vaults as subdirectories |
WRITE_VAULT |
Claude |
Name of the only vault where write operations are permitted |
Example layout expected under HOST_VAULT_PATH:
/path/to/your/vaults/
Claude/ ← write operations land here
Work/ ← readable but not writable
Personal/ ← readable but not writable
Usage
As a standalone stdio server
# via the installed CLI entry point
HOST_VAULT_PATH=/path/to/your/vaults obsidian-mcp-guard
# or via python -m
HOST_VAULT_PATH=/path/to/your/vaults python -m obsidian_mcp_guard
Claude Desktop / Cursor config
{
"mcpServers": {
"obsidian": {
"command": "uvx",
"args": ["obsidian-mcp-guard"],
"env": {
"HOST_VAULT_PATH": "/path/to/your/vaults",
"WRITE_VAULT": "Claude"
}
}
}
}
Claude Code
claude mcp add obsidian -- uvx obsidian-mcp-guard
Pass environment variables with -e:
claude mcp add obsidian -e HOST_VAULT_PATH=/path/to/your/vaults -e WRITE_VAULT=Claude -- uvx obsidian-mcp-guard
Mounted into another FastMCP server
from contextlib import asynccontextmanager
from fastmcp import FastMCP
from obsidian_mcp_guard import create_vault_server
@asynccontextmanager
async def lifespan(app):
await app.import_server(create_vault_server(
vault_path="/path/to/your/vaults",
write_vault="Claude"
))
yield
mcp = FastMCP("my-agent", lifespan=lifespan)
@mcp.tool()
def search_notes(...):
...
Tools
| Tool | Description |
|---|---|
read_note(source) |
Return full content of a note in vault/path.md format |
list_notes(vault, folder?, recursive?) |
List note paths within a vault or subfolder |
create_note(source, content, overwrite?) |
Create a note; blocked by lint errors |
update_note(source, content, mode?) |
Overwrite or append to a note; blocked by lint errors |
delete_note(source) |
Move a note to .trash/ (recoverable) |
move_note(source_path, dest_path, create_dirs?) |
Move a note within the write vault; rewrites wikilinks in all vault files |
lint_note(content) |
Pre-validate content without writing; returns {valid, errors, warnings} |
Development
make install # install package + test dependencies
make test # run tests with coverage (90% minimum)
make build # build source and wheel distributions
make clean # remove build artefacts and cache files
See CONTRIBUTING.md for full guidelines.
Related projects
- mdlint-obsidian — the lint engine used to validate markdown against Obsidian's rendering rules
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.