read-docs-mcp

What is read-docs-mcp?

read-docs-mcp is a Model Context Protocol (MCP) server that enables AI agents to access and understand package documentation through a structured interface. It supports two modes: Read Documentation Mode, which reads docs from a given git repository, and Create Documentation Mode, which guides users in setting up documentation files. It is intended for developers who want AI assistants to explore or create documentation for their libraries.

How to use read-docs-mcp?

read-docs-mcp is invoked via npx -y read-docs-mcp with optional command-line arguments such as --name, --git-repo-path, --branch, --docs-path, and --clone-location. It can be configured in Cursor by adding a JSON block to the Cursor settings under mcpServers. For Read Documentation Mode, provide both --name and --git-repo-path; omit them to use Create Documentation Mode.

Key features of read-docs-mcp

• Dual usage modes: read existing documentation or create documentation structure
• Automatically generates MCP tools from documentation structure
• Supports multiple documentation modules (hooks, components, utilities, etc.)
• Configurable naming patterns for documentation files and module folders
• Provides listing, overview, and detailed documentation access per module
• Falls back to package.json for version information when not specified in config

Use cases of read-docs-mcp

• Explore available modules and hooks of a library through an AI agent
• Retrieve detailed documentation for a specific component or utility
• Guide the creation of a documentation folder structure and configuration files
• Learn how to use individual features (e.g., authentication, form validation) from documentation
• Set up documentation for a library being developed, with step-by-step instructions

FAQ from read-docs-mcp

What is the difference between Read Documentation Mode and Create Documentation Mode?

In Read Documentation Mode (invoked with --name and --git-repo-path), the server reads documentation from a specified repository and generates tools to access it. In Create Documentation Mode (no repository info provided), the server provides instructions for setting up documentation files.

How do I handle authentication for private repositories?

read-docs-mcp requires direct cloning of the target repository. For SSH URLs, ensure SSH keys are configured. For HTTPS URLs, use a personal access token in the URL or configure Git credential storage on your machine.

What command-line arguments are available?

The server supports --name (package name), --git-repo-path (HTTP or SSH URL), --branch (default: main), --docs-path (default: docs), and --clone-location (default: ~/.temp-repo). --name and --git-repo-path are required for Read Documentation Mode.

How do I set up the documentation structure for my library?

Create a read-docs-mcp.json file in the docs directory with fields like name, description, moduleList, and moduleFolderNamingPattern. For each module, create a folder and a read-module-docs-mcp.json file defining tools and file names. Then add markdown files (list.md, overview.md, and individual detail files).

What naming patterns are supported for module folders and detail files?

The supported patterns are kebab-case (default), camelCase, snake_case, PascalCase, and original (uses names as-is). These are configured via moduleFolderNamingPattern in the main configuration file.