deepwiki-dl

Download deepwiki.com content as separate Markdown files using the deepwiki MCP server.

Note: This is an unofficial tool. The deepwiki MCP server's behavior may change at any time, which could break this tool's functionality.

Why?

The deepwiki MCP server's read_wiki_contents tool returns all pages concatenated together with # Page: [title] delimiters (without newlines before them). This bursts coding AI's context usage.

This tool splits the combined content into separate, properly numbered Markdown files based on the structure, making it easier to:

• Read only the parts you (the AI) want
• Minimize context usage

Installation

npm install -g @heguro/deepwiki-dl
# or use with npx
npx @heguro/deepwiki-dl owner/repo [outDir]

Usage

deepwiki-dl owner/repo [outDir]

Arguments

• owner/repo - GitHub repository (e.g., modelcontextprotocol/typescript-sdk)
• outDir - Optional output directory. Defaults to {repo-name}-deepwiki/

Examples

Download documentation for the MCP TypeScript SDK:

deepwiki-dl modelcontextprotocol/typescript-sdk

This will create a directory typescript-sdk-deepwiki/ with:

• _wiki_structure.md - The structure from read_wiki_structure
• 1 Overview.md - Individual pages with proper numbering
• 1.1 Installation and Setup.md
• 1.2 Core Concepts.md
• 2 Architecture.md
• etc.

Specify a custom output directory:

deepwiki-dl modelcontextprotocol/typescript-sdk ./my-docs

How it works

This tool connects to the deepwiki MCP server at https://mcp.deepwiki.com/mcp and:

1. Calls read_wiki_structure to get the table of contents with numbered sections
2. Calls read_wiki_contents to get all page content in a single string
3. Parses the content which comes in format # Page: [title]\n\n[content]
4. Matches page titles with the structure to determine proper numbering
5. Saves each page as a separate Markdown file with format N [title].md or N.N [title].md

Development

# Install dependencies
pnpm install

# Build
pnpm run build

# Lint
pnpm run lint

# Format
pnpm run format

License

CC0-1.0 - Public Domain