Grokipedia 内容检索 MCP 服务器

Grokipedia MCP Server

MCP server for searching and retrieving content from Grokipedia

The User of the MCP assumes full responsibility for interacting with Grokipedia.

Please see the Xai Terms of Service if you have any doubts.

Elon, please don't sue me. I only wanted my agents to have access to truthful information and stop referencing wikipedia all the time.

Quick Start

Add this to your MCP configuration file:

{
  "mcpServers": {
    "grokipedia": {
      "command": "uvx",
      "args": ["grokipedia-mcp"]
    }
  }
}

Verifying Installation

You should see the Grokipedia server available with these tools:

• search - Search with filters
• get_page - Get page overview
• get_page_content - Get full content
• get_page_citations - Get citations
• get_linked_pages - Get pages linked from an article
• get_page_sections - List all section headers
• get_page_section - Extract specific sections

And these prompts:

• research_topic - Research workflow
• find_sources - Find citations
• explore_related - Explore connections
• compare_topics - Compare two topics

Features

• Search with Filters: Search with sorting (relevance/views) and filtering (min views)
• Page Content: Retrieve articles, citations, and metadata with smart truncation
• Related Pages: Discover linked/related articles
• Section Extraction: Get specific sections from long articles
• Smart Suggestions: Helpful alternatives when pages aren't found
• Guided Prompts: Pre-built workflows for research, sources, exploration

Installation (Development)

Using uv:

cd grokipedia-mcp
uv sync

For development with MCP Inspector and CLI tools:

uv sync --dev

Usage

Run with MCP Inspector (Development)

The fastest way to test and debug (requires dev dependencies):

uv run --dev mcp dev main.py

This launches the MCP Inspector UI where you can:

• Explore available tools
• Test search queries
• Retrieve page content
• View structured output

Run Directly

# Using the installed entry point
uv run grokipedia-mcp

# Or as a Python module
uv run python -m grokipedia_mcp

# Or directly
uv run python main.py

Available Tools

search

Search for articles in Grokipedia with filtering and sorting options.

Parameters:

• query (string, required) - Search query
• limit (int, optional, default: 12) - Maximum number of results
• offset (int, optional, default: 0) - Pagination offset
• sort_by (string, optional, default: "relevance") - Sort by "relevance" or "views"
• min_views (int, optional) - Filter to articles with at least this many views

Returns: List of search results with title, slug, snippet, relevance score, and view count.

Examples:

// Basic search
{"query": "machine learning", "limit": 5}

// Sort by most viewed
{"query": "python", "sort_by": "views"}

// Filter popular articles only
{"query": "artificial intelligence", "min_views": 1000}

get_page

Get complete page information including metadata, content preview, and citations summary. Includes smart suggestion of alternatives if page not found.

Parameters:

• slug (string, required) - Article identifier (from search results)
• max_content_length (int, optional, default: 5000) - Maximum content length

Returns: Complete page object with metadata, truncated content, and citation summaries.

Features:

• Suggests similar pages if the requested slug doesn't exist
• Provides overview with content preview and citations

Use this when: You need an overview of a page with metadata and a content preview.

Example:

{"slug": "Machine_learning"}

get_page_content

Get only the article content without citations or metadata.

Parameters:

• slug (string, required) - Article identifier
• max_length (int, optional, default: 10000) - Maximum content length

Returns: Only the article content (title and content text).

Use this when: You need to read the full article content without citations.

Example:

{"slug": "Machine_learning", "max_length": 15000}

get_page_citations

Get the citations list for a specific page.

Parameters:

• slug (string, required) - Article identifier
• limit (int, optional) - Maximum number of citations to return (returns all if not specified)

Returns: List of citations with titles, URLs, and descriptions. Includes total count and returned count.

Use this when: You need to access source references and citations.

Examples:

// Get all citations
{"slug": "Machine_learning"}

// Get first 10 citations only
{"slug": "Machine_learning", "limit": 10}

get_linked_pages

Get pages that are linked from a specific article. Grokipedia no longer publishes a curated linked-pages field, so links are derived from /page/<slug> markdown links in the article content and ranked by mention frequency, then first position.

Parameters:

• slug (string, required) - Article identifier
• limit (int, optional, default: 10) - Maximum number of linked pages to return

Returns: List of linked pages with titles, slugs, and mention counts.

Use this when: You want to discover topics an article references or explore connections between articles.

Examples:

// Get linked pages
{"slug": "Machine_learning"}

// Get more linked pages
{"slug": "Quantum_computing", "limit": 20}

get_page_sections

Get a list of all section headers in an article.

Parameters:

• slug (string, required) - Article identifier

Returns: List of all section headers with their levels (h1, h2, h3, etc.).

Use this when: You want to see the structure/outline of an article before reading specific sections.

Example:

{"slug": "Machine_learning"}

get_page_section

Extract a specific section from an article by header name.

Parameters:

• slug (string, required) - Article identifier
• section_header (string, required) - Section header to extract (case-insensitive)
• max_length (int, optional, default: 5000) - Maximum section content length

Returns: Content of the specified section only.

Use this when: You need just one section of a long article (e.g., "Applications", "History", "Examples").

Examples:

// Get specific section
{"slug": "Neural_networks", "section_header": "Applications"}

// Get longer section
{"slug": "Python", "section_header": "Syntax", "max_length": 10000}

Note: Articles can be 100,000+ characters. Content is automatically truncated to prevent overwhelming LLM context windows. Use the max_length parameters to control the amount returned.

Prompts

The server provides pre-built prompts for common workflows:

research_topic

Guided workflow to research a topic: search → retrieve → analyze related pages and citations

find_sources

Find authoritative sources and citations for academic/research purposes

explore_related

Discover connections between topics and suggested further reading

compare_topics

Compare two topics side-by-side with their content and citations

Architecture

The server uses:

• FastMCP for declarative MCP server implementation
• grokipedia-api-sdk AsyncClient for API communication
• Lifespan context for client connection management
• Structured output using Pydantic models from the SDK
• Comprehensive error handling with specific exception types

Error Handling

The server handles various error scenarios:

• ValueError for invalid parameters or not found pages
• RuntimeError for network or API errors
• Detailed logging at debug, info, warning, and error levels

Development

Project Structure

grokipedia-mcp/
├── grokipedia_mcp/
│   ├── __init__.py       # Package exports
│   ├── __main__.py       # CLI entry point
│   └── server.py         # FastMCP server implementation
├── main.py               # Direct execution entry point
├── pyproject.toml        # Project configuration
└── README.md             # This file

Testing

Use the MCP Inspector for interactive testing:

uv run mcp dev main.py

License

MIT