# KatanA Markdown Linter [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/HiroyukiFuruno/katana-markdown-linter  
**GitHub Stars:** 0  
**npm Downloads (last month):** 1165  
**Views:** 5  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/katana-markdown-linter

## Description
Markdown linter and workspace-scoped stdio MCP server.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "katana-markdown-linter": {
    "command": "npx",
    "args": ["--yes","katana-markdown-linter@0.20.1"]
  }
}
```

## Documentation

## What KatanA Markdown Linter does

KatanA Markdown Linter brings the project's Markdown analysis engine to MCP-compatible agents and tools. It is designed for technical writing, specification documents, and documentation workflows where an agent needs machine-readable lint results or controlled file changes.

The underlying engine follows the markdownlint rule model and supports the active upstream rule catalog. Its capabilities cover three related tasks: checking Markdown for rule violations, applying fixes where changes are considered safe, and normalizing layout through formatter behavior. The project also provides a standalone CLI, Rust library, editor server, and distribution packages, but the MCP component is the workspace-oriented interface for agent use.

## How it works

The MCP server communicates over standard input and output and is scoped to a workspace. Start it with the `kml-mcp` executable and provide an absolute workspace root when needed. Agents can then request structured Markdown lint access against files in that workspace rather than relying on shell output parsing.

The same engine supports safe lint-driven fixes and a formatter for indentation and newline normalization. Safe fixes are intended for low-risk violations; the broader CLI separately exposes explicit opt-in handling for unsafe fixes. This separation makes the MCP server suitable for workflows that need diagnostics first and conservative rewriting afterward.

The project supports localized rule metadata and descriptions for English, Japanese, Simplified Chinese, Traditional Chinese, Korean, Portuguese, French, German, Spanish, and Italian. That localization is part of the underlying engine and can help applications present rule information in a user-facing language.

## Setup and configuration

Install the package through Cargo, npm, PyPI, GitHub Releases, or Homebrew. Cargo installs the Rust package and CLI; npm and PyPI provide launchers that obtain the matching release archive on first use. The npm package exposes `kml-mcp`, and the documented launch pattern supplies `--workspace-root` with an absolute path.

For example, an npm-based setup can run the MCP process with `npx --yes katana-markdown-linter@latest kml-mcp --workspace-root /absolute/path/to/workspace`. A globally installed package can invoke the same `kml-mcp` executable directly. Configure the command and its arguments in the MCP client that launches stdio servers.

The repository also supports the `kml` CLI through Cargo, npm, PyPI, standalone release archives, and Homebrew. This can be useful for validating the same lint behavior in local scripts or CI alongside agent-driven checks. The project is licensed under MIT.

## Tools and capabilities

- Markdownlint-compatible rule checks.
- Structured lint access through MCP.
- Safe fixes for low-risk violations.
- Formatting for indentation and newline normalization.
- Workspace-oriented operation through a stdio process.
- Localized rule metadata and descriptions across multiple languages.

The broader `kml` toolchain also supports JSON output, stdin processing, ignore handling, statistics, configuration schema access, and rule documentation. These features belong to the project’s CLI or library surfaces; use the MCP server when the primary requirement is exposing Markdown analysis to an agent.

_Full upstream README: https://allmcps.com/mcp/katana-markdown-linter/readme_

