LLM API Documentation Collection

Aggregated API documentation from 10 major LLM providers

English • 简体中文

🌐 Live Demo

---

✨ Features

• 📚 1600+ Documents - Comprehensive API documentation from major LLM providers
• 🌍 Multi-language - Support for 12 languages including English, Chinese, Japanese, Korean, etc.
• 🔍 Full-text Search - Search across all documentation with instant results
• 🤖 MCP Integration - Let AI assistants read all docs via Model Context Protocol
• 🎨 Modern UI - Clean, responsive design with dark/light theme
• 📦 Batch Download - Download all docs as ZIP files

📋 Supported Vendors

Vendor	Documents	Languages	Crawl Type
OpenAI	51	English	Manual
Anthropic Claude	963	12 languages	Auto
Google Gemini	67	Chinese	Auto
Meta Llama	22	English	Manual
xAI Grok	59	English	Auto
Moonshot Kimi	72	Chinese, English	Manual
Zhipu BigModel	242	Chinese, English	Mixed
MiniMax	84	Chinese, English	Auto
MegaLLM	120	English, Chinese, Russian	Auto
DeepSeek	61	Chinese, English	Manual

🚀 Quick Start

View Documentation Website

# Build the docs index
python3 build_docs_site.py

# Start local server
cd docs-site && python3 -m http.server 8080

# Open http://localhost:8080 in browser

Download Documentation

# Install dependencies
pip install requests

# Run batch download
python3 batch_download_docs.py

MCP Server Integration

# Install MCP SDK
pip install mcp

# Add to your MCP client config

{
  "mcpServers": {
    "llm-docs": {
      "command": "python3",
      "args": ["mcp_server.py"],
      "disabled": false
    }
  }
}

📁 Project Structure

├── docs-site/              # Static documentation website
│   ├── index.html
│   ├── styles.css
│   ├── app.js
│   └── logos/              # Vendor logos
├── OpenAI/                 # OpenAI docs
├── Anthropic Claude/       # Claude docs (12 languages)
├── Google Gemini/          # Gemini docs
├── Meta Llama/             # Llama docs
├── X Grok/                 # Grok docs
├── Moonshot Kimi/          # Kimi docs
├── BigModel Zhipu/         # Zhipu docs
├── MiniMax/                # MiniMax docs
├── MegaLLM/                # MegaLLM docs
├── DeepSeek/               # DeepSeek docs
├── mcp_server.py           # MCP server
├── build_docs_site.py      # Site builder
└── download_*.py           # Crawler scripts

---

📖 Technical Documentation

Google Gemini

URL Pattern: https://ai.google.dev/gemini-api/docs/{path}.md.txt?hl=zh-cn

Findings:

• Google docs support .md.txt suffix to get Markdown format
• Use ?hl=zh-cn parameter for Chinese version
• Crawled 70 document links from HTML page sidebar
• Downloaded content may have <br /> tags at the beginning, needs cleaning
• Multi-language: Machine translation (Google Cloud Translation API), pages marked "translated by Google", only need to download one language

Download Result: 67/70 successful (3 404s: nanobanana, pricing, partner-integration)

Directory Structure:

Google Gemini/docs/
├── 01-开始使用/     (5 docs)
├── 02-模型/         (8 docs)
├── 03-核心功能/     (12 docs)
├── 04-工具和代理/   (8 docs)
├── 05-Live API/     (5 docs)
├── 06-指南/         (15 docs)
├── 07-资源/         (12 docs)
└── 08-政策/         (2 docs)

---

Anthropic Claude

URL Pattern: https://platform.claude.com/docs/{locale}/{path}.md

Findings:

• Claude docs is a Next.js app, supports .md suffix to get Markdown
• Supports 12 languages: en, de, es, fr, it, ja, ko, pt-BR, ru, zh-CN, zh-TW, id
• ⚠️ Important: Language codes are case-sensitive! zh-CN is correct, zh-cn returns HTML 404 page
• Extracted 85 document links from HTML page
• Returned Markdown contains some JSX component tags (e.g., <DocsSearchBar />)
• Multi-language: Native multi-language, each language maintained independently, content may differ, recommend downloading all

Download Result: 963 documents (some languages have fewer docs, e.g., Français 84, Italiano 30)

Directory Structure:

Anthropic Claude/
├── English/
│   ├── home.md
│   ├── about-claude/
│   │   ├── models/
│   │   ├── pricing.md
│   │   └── model-deprecations.md
│   ├── build-with-claude/
│   │   ├── prompt-engineering/
│   │   └── ...
│   ├── agents-and-tools/
│   ├── agent-sdk/
│   ├── test-and-evaluate/
│   └── release-notes/
├── 简体中文/
├── 日本語/
└── ... (other languages)

---

Moonshot Kimi

URL Pattern: https://platform.moonshot.cn/docs/{path} (Chinese) / https://platform.moonshot.cn/en-US/docs/{path} (English)

Findings:

• Moonshot is a Next.js SPA, content rendered by JavaScript, cannot directly get Markdown
• Need to use Playwright to render page, extract content, then convert to Markdown with markdownify
• Supports 2 languages: zh-CN (Simplified Chinese), en-US (English)
• Uses Nextra documentation framework, well-structured HTML, good conversion results
• Multi-language: Native multi-language, Chinese and English maintained independently

Download Result: 36 pages × 2 languages = 72 documents

Dependencies: pip install playwright markdownify && playwright install chromium

---

X Grok

URL Pattern: https://docs.x.ai/llms{path}.md

Findings:

• Document list available from https://docs.x.ai/llms.txt
• Requires User-Agent and Referer headers, otherwise returns 403
• Returned Markdown has ===/docs/xxx=== markers at the beginning, needs cleaning
• Multi-language: English only, no multi-language support

Download Result: 59 documents

---

OpenAI

URL Pattern: No direct Markdown endpoint, need to render page and extract

Findings:

• OpenAI docs is a SPA with Cloudflare bot protection
• Official "Copy page" button can copy Markdown, but cannot directly request it
• Need to use undetected-chromedriver to bypass bot protection
• Multi-language: English only, no multi-language support

Download Result: 51/54 successful (3 model pages failed)

Dependencies: pip install undetected-chromedriver selenium markdownify

---

Zhipu BigModel

Zhipu has two documentation sites with different content:

Chinese Site (docs.bigmodel.cn) - Mintlify Framework

URL Pattern: https://docs.bigmodel.cn/cn/{section}/{path}.md

• Uses Mintlify documentation framework, supports .md suffix to get Markdown directly
• Download Result: 129 documents

English Site (open.bigmodel.cn) - Vue SPA

• Vue SPA, content rendered by JavaScript
• Need to use Playwright to render page
• Download Result: 113 documents

Dependencies: pip install playwright markdownify && playwright install chromium

---

MiniMax

URL Pattern:

• English site: https://platform.minimax.io/docs/{path}.md
• Chinese site: https://platform.minimaxi.com/docs/{path}.md

Findings:

• Uses Mintlify documentation framework, supports .md suffix to get Markdown directly
• Chinese and English are two independent sites with different domains

Download Result: English 37 + Chinese 47 = 84 documents

---

Meta Llama

URL Pattern: No direct Markdown endpoint, need to render page and extract

Findings:

• React SPA (Facebook style), content rendered by JavaScript
• Need to use Playwright to render page

Download Result: 22 documents

Dependencies: pip install playwright markdownify && playwright install chromium

---

MegaLLM

URL Pattern: https://docs.megallm.io/{lang}/{path}.md

Findings:

• Provides llms.txt file with complete document list: https://docs.megallm.io/llms.txt
• Supports .md suffix to get Markdown directly
• An aggregation platform connecting 70+ LLMs with unified API
• Multi-language: Supports 3 languages: en (English), cn (Chinese), ru (Russian)

Download Result: 40 × 3 = 120 documents

---

DeepSeek

URL Pattern: https://api-docs.deepseek.com/{path} (English) / https://api-docs.deepseek.com/zh-cn/{path} (Chinese)

Findings:

• Uses Docusaurus documentation framework
• Need to use Selenium to render page

Download Result: 61 documents (31 English + 30 Chinese)

Dependencies: pip install selenium html2text

---

🛠️ Dependencies

Script	Dependencies
Basic crawlers	requests
Moonshot, Meta, Zhipu EN	playwright markdownify
OpenAI	undetected-chromedriver selenium markdownify
DeepSeek	selenium html2text

📝 File Description

File	Description
batch_download_docs.py	Gemini + Claude download script
download_moonshot.py	Moonshot download script
download_grok.py	X Grok download script
download_openai_uc.py	OpenAI download script
download_zhipu.py	Zhipu Chinese site download script
download_zhipu_en.py	Zhipu English site download script
download_minimax.py	MiniMax download script
download_meta.py	Meta Llama download script
download_megallm.py	MegaLLM download script
download_deepseek.py	DeepSeek download script
build_docs_site.py	Documentation site builder
mcp_server.py	MCP server for AI assistants

---

🌐 Documentation Website

A static documentation website is included in the docs-site/ folder:

python3 build_docs_site.py
cd docs-site && python3 -m http.server 8080

Features:

• 10 LLM vendor documentation aggregated
• Multi-language document tabs
• Site language switcher (English/Chinese/Japanese)
• Dark/Light theme toggle
• Code block copy buttons
• Category navigation in sidebar
• Modern responsive design

---

🤖 MCP Server

Available Tools

Tool	Description
list_vendors	List all LLM vendors with doc counts
list_docs	List all documents for a specific vendor
read_doc	Read full content of a document
search_docs	Search across all documentation
get_doc_stats	Get collection statistics

Configuration

{
  "mcpServers": {
    "llm-docs": {
      "command": "python3",
      "args": ["mcp_server.py"],
      "disabled": false,
      "autoApprove": ["list_vendors", "list_docs", "read_doc", "search_docs", "get_doc_stats"]
    }
  }
}

Resource URIs

• llmdocs://openai - Get OpenAI vendor info and doc list
• llmdocs://anthropic/en/about-claude/pricing.md - Read specific document

---

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

⚠️ Disclaimer

This project is for educational and research purposes only.

• All documentation content belongs to their respective owners (OpenAI, Anthropic, Google, Meta, xAI, Moonshot, Zhipu, MiniMax, MegaLLM, DeepSeek)
• This project does not claim ownership of any documentation content
• The crawled documentation is intended for personal learning and development reference
• Please respect the terms of service of each platform
• Do not use this project for commercial purposes without proper authorization
• The maintainers are not responsible for any misuse of this project

If you are a representative of any of the included platforms and have concerns about this project, please open an issue and we will address it promptly.

🙏 Acknowledgments

• Thanks to all LLM providers for their excellent documentation
• Built with ❤️ for the developer community