Search MCP Server

MCP server providing 27 tools for web search, content extraction, and data processing. No API keys required.

Built with a modular registry architecture for dynamic tool loading, categorization, and discovery.

Installation

From PyPI (recommended)

pip install mcp-search-server

From source

git clone https://github.com/KazKozDev/mcp-search-server.git
cd mcp-search-server
pip install -e .

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "search": {
      "command": "mcp-search-server"
    }
  }
}

Available Tools (27)

Meta Tools - Tool Discovery

Why LLMs need this: With 27 tools available, finding the right one for a task is challenging. This meta-tool enables LLMs to search through available tools by keywords, categories, or use cases.

What it does: Searches the tool registry by name, description, category, or tags. Returns ranked results with relevance scores.

Parameters:

• query (required): Search query (e.g., "find web pages", "date time", "github")
• limit: Max results (default: 5)

Example:

{"query": "search academic papers", "limit": 3}

Why LLMs need this: Tools are organized into categories (web, knowledge, social, etc.). This tool helps discover what categories exist and how many tools are in each.

What it does: Returns all available tool categories with counts.

Example:

{}

Why LLMs need this: Before using a tool, LLMs need to understand its parameters, return format, and usage examples. This tool provides complete documentation for any registered tool.

What it does: Returns detailed metadata for a specific tool including description, parameters, category, priority, and usage info.

Parameters:

• tool_name (required): Name of the tool

Example:

{"tool_name": "search_arxiv"}

Web Search & Content

Why LLMs need this: LLMs have a knowledge cutoff date and cannot access current information. This tool gives them real-time access to the web, enabling answers about recent events, current prices, latest news, and up-to-date documentation.

What it does: Searches the web using DuckDuckGo with support for web and news modes, time filtering, and region-specific results.

Parameters:

• query (required): Search query
• limit: Max results (default: 10)
• mode: "web" or "news"
• timelimit: "d" (day), "w" (week), "m" (month), "y" (year)

Example:

{"query": "Python async programming", "limit": 5, "mode": "news", "timelimit": "w"}

Why LLMs need this: LLMs cannot look up real addresses, business locations, or geographic data. This tool enables location-based queries like finding nearby businesses, getting addresses, and working with geographic coordinates.

What it does: Searches for places, addresses, and locations with geographic coordinates. Returns structured location data including names, addresses, and coordinates.

Parameters:

• query (required): Place or address to search
• limit: Max results (default: 5)

Example:

{"query": "coffee shops near Times Square NYC", "limit": 3}

Why LLMs need this: Raw HTML is cluttered with navigation, ads, and scripts. LLMs need clean, readable text to understand and summarize web content effectively. This tool extracts just the main content.

What it does: Extracts readable content from any webpage, removing ads, navigation, and boilerplate. Uses multiple parsing methods (Readability, Newspaper3k, BeautifulSoup) with automatic fallback.

Parameters:

• url (required): URL to extract content from

Example:

{"url": "https://example.com/article"}

Why LLMs need this: PDFs are a common format for research papers, reports, and documentation, but LLMs cannot read binary files directly. This tool converts PDF content to text for analysis.

What it does: Downloads and extracts text content from PDF documents using PyPDF2 or pdfplumber with automatic library selection.

Parameters:

• url (required): PDF URL
• max_chars: Maximum characters to extract (default: 50000)

Example:

{"url": "https://arxiv.org/pdf/2303.08774.pdf", "max_chars": 100000}

Wikipedia

Why LLMs need this: Wikipedia is a reliable, structured knowledge base. While LLMs have some Wikipedia knowledge from training, this tool provides access to the latest content and helps find specific articles on any topic.

What it does: Searches Wikipedia for articles matching a query. Returns article titles, snippets, and URLs.

Parameters:

• query (required): Search query
• limit: Max results (default: 5)

Example:

{"query": "machine learning", "limit": 3}

Why LLMs need this: Quick factual lookups need concise, authoritative information. Wikipedia summaries provide verified facts without the overhead of full articles, perfect for answering specific questions.

What it does: Gets a concise summary of a Wikipedia article - the introductory paragraphs that define the topic.

Parameters:

• title (required): Article title

Example:

{"title": "Artificial Intelligence"}

Why LLMs need this: For in-depth research, LLMs need complete information with all sections, references, and details. This tool provides the full Wikipedia article for comprehensive analysis.

What it does: Gets the complete content of a Wikipedia article with all sections, supporting multiple languages.

Parameters:

• title (required): Article title
• lang: Language code (default: "en")

Example:

{"title": "Quantum computing", "lang": "en"}

Academic & Research

Why LLMs need this: Scientific research moves fast. arXiv hosts the latest preprints in physics, math, CS, and more. This tool gives LLMs access to cutting-edge research before it's even published in journals.

What it does: Searches arXiv for academic papers with filtering by category. Returns titles, abstracts, authors, and links.

Parameters:

• query (required): Search query
• max_results: Max results (default: 10)
• category: arXiv category (e.g., "cs.AI", "physics")

Example:

{"query": "transformer neural networks", "max_results": 5, "category": "cs.AI"}

Why LLMs need this: Medical and health questions require peer-reviewed sources. PubMed is the authoritative database for biomedical research, enabling LLMs to cite real studies rather than general knowledge.

What it does: Searches PubMed for biomedical and life science research papers. Returns titles, abstracts, authors, DOIs, and publication info.

Parameters:

• query (required): Search query
• max_results: Max results (default: 10)

Example:

{"query": "CRISPR gene therapy", "max_results": 5}

Why LLMs need this: For current events and global news analysis, LLMs need access to worldwide media coverage. GDELT monitors news from every country, enabling analysis of how stories develop globally.

What it does: Searches the GDELT database for global news articles and events with time filtering.

Parameters:

• query (required): Search query
• timespan: "1d", "7d", or "1m" (default: "1d")
• max_results: Max results (default: 10)

Example:

{"query": "climate summit", "timespan": "7d", "max_results": 5}

GitHub

Why LLMs need this: Developers ask about libraries, frameworks, and code examples. This tool helps LLMs find relevant repositories, compare alternatives, and recommend tools based on real popularity metrics (stars, forks).

What it does: Searches GitHub for repositories by keywords with sorting by stars, forks, or update date.

Parameters:

• query (required): Search query
• max_results: Max results (default: 5)
• sort: "stars", "forks", "updated" (default: "stars")

Example:

{"query": "python web framework", "max_results": 10, "sort": "stars"}

Why LLMs need this: README files contain installation instructions, API docs, and usage examples. This tool lets LLMs read actual documentation to provide accurate, up-to-date guidance on using any library.

What it does: Fetches the README file content from a GitHub repository in markdown format.

Parameters:

• repo (required): Repository in "owner/repo" format

Example:

{"repo": "langchain-ai/langchain"}

Reddit

Why LLMs need this: Reddit contains real user experiences, reviews, and discussions. For questions like "what's the best X" or "how do people feel about Y", Reddit provides authentic community perspectives that formal sources lack.

What it does: Searches Reddit for posts across all subreddits or within a specific one, with time filtering.

Parameters:

• query (required): Search query
• subreddit: Specific subreddit (optional)
• limit: Max results (default: 10)
• time_filter: "hour", "day", "week", "month", "year", "all"

Example:

{"query": "best programming languages 2024", "subreddit": "programming", "limit": 5}

Why LLMs need this: The real value of Reddit is in the comments - detailed explanations, counterarguments, and community voting. This tool extracts comment threads for deeper analysis of discussions.

What it does: Fetches comments from a specific Reddit post with scores and hierarchy.

Parameters:

• url (required): Reddit post URL
• limit: Max comments (default: 10)

Example:

{"url": "https://www.reddit.com/r/Python/comments/abc123/post_title/", "limit": 20}

Date, Time & Location

Why LLMs need this: LLMs have no concept of "now" - they don't know the current date or time. This tool enables time-aware responses: scheduling, deadlines, "is it open now?", age calculations, and timezone conversions.

What it does: Gets the current date and time for any timezone with detailed components (day of week, week number, Unix timestamp).

Parameters:

• timezone: Timezone name (default: "UTC")
• include_details: Include additional info (default: true)

Example:

{"timezone": "America/New_York", "include_details": true}

Returns: ISO datetime, date/time components, day of week, week number, Unix timestamp.

Why LLMs need this: LLMs have no concept of "here" - they don't know where the user is. This tool enables location-aware responses: local weather, nearby services, correct timezone, and region-specific information.

What it does: Gets geographic location from an IP address including country, city, timezone, coordinates, and ISP info.

Parameters:

• ip_address: IP to lookup (optional, uses server IP if omitted)

Example:

{"ip_address": "8.8.8.8"}

Returns: Country, region, city, ZIP, timezone, lat/lon, ISP, AS number.

Analysis & Processing

Why LLMs need this: Not all sources are equal. When citing information, LLMs need to distinguish between peer-reviewed research, reputable news, and random blogs. This tool provides objective credibility metrics to support source evaluation.

What it does: Assesses web source credibility using 30+ signals including domain age (WHOIS), citation network (PageRank), and content analysis with Bayesian confidence intervals.

Parameters:

• url (required): URL to assess
• title: Document title (optional)
• content: Full text content (optional, improves accuracy)
• metadata: Additional metadata (year, authors, citations, doi, is_peer_reviewed)

Example:

{"url": "https://arxiv.org/abs/2301.00234", "metadata": {"is_peer_reviewed": true}}

Returns: Credibility score (0-1), confidence interval, category, PageRank, 30+ signal scores.

Why LLMs need this: Long documents exceed context limits. Before analyzing a large PDF or article, this tool can create a concise summary, letting LLMs work with more sources without running out of context.

What it does: Summarizes long text using TF-IDF extraction, keyword-based selection, or fast heuristic methods. Runs locally with no external APIs.

Parameters:

• text (required): Text to summarize
• strategy: "auto", "extractive_tfidf", "extractive_keyword", "heuristic"
• compression_ratio: Target ratio 0.1-0.9 (default: 0.3)

Example:

{"text": "Long article text here...", "strategy": "extractive_tfidf", "compression_ratio": 0.3}

Returns: Summary, method used, statistics (original/summary length, compression ratio).

Why LLMs need this: LLMs are notoriously bad at math. Even simple arithmetic can produce wrong answers. This tool provides accurate calculations for everything from basic math to trigonometry and logarithms.

What it does: Performs mathematical calculations using safe AST parsing (no eval). Supports arithmetic, trigonometry, logarithms, factorials, and mathematical constants.

Parameters:

• expression (required): Math expression

Supported: +, -, *, /, **, ^, %, //, sqrt, sin, cos, tan, log, log10, exp, factorial, pi, e, and more.

Example:

{"expression": "sqrt(144) + sin(pi/2) * 10"}

File Management

Why LLMs need this: Users often want to discuss files on their system. This tool lets LLMs read text, PDFs, Word docs, and more to answer questions about file contents, analyze data, or help with editing.

What it does: Reads content from text, PDF, Word, Excel, or image files with automatic format detection.

Parameters:

• path (required): File path (relative paths use data/files/ as base)

Example:

{"path": "notes.txt"}

Why LLMs need this: LLMs can generate code, documents, and data, but without file writing they can only display output. This tool lets them save generated content to actual files users can use.

What it does: Writes content to a file, creating it if it doesn't exist. UTF-8 text files only.

Parameters:

• path (required): File path
• content (required): Content to write

Example:

{"path": "output.txt", "content": "Hello, World!"}

Why LLMs need this: For logging, note-taking, and incremental data collection, appending is safer than overwriting. This tool adds content to existing files without losing previous data.

What it does: Appends content to an existing file or creates a new one.

Parameters:

• path (required): File path
• content (required): Content to append

Example:

{"path": "log.txt", "content": "\nNew log entry"}

Why LLMs need this: Before reading or writing files, LLMs need to know what exists. This tool provides directory listings with file sizes and types, enabling file management workflows.

What it does: Lists files and directories with sizes and types.

Parameters:

• path: Directory path (default: data/files/)

Example:

{"path": ""}

Why LLMs need this: Complete file management requires deletion. This tool enables cleanup of temporary files, old outputs, and user-requested deletions with security restrictions.

What it does: Deletes a file (restricted to data/files/ directory for security).

Parameters:

• path (required): File path to delete

Example:

{"path": "temp.txt"}

Architecture

The server uses a modular registry architecture:

• Dynamic Tool Loading: Tools are registered in a central registry with metadata (category, priority, tags)
• Category-Based Organization: Tools organized into 7 categories: meta, web, knowledge, social, analysis, context, files
• Meta-Tool Discovery: Built-in tools (search_tools, list_tool_categories, get_tool_info) help LLMs discover and understand available capabilities
• Extensible: Easy to add new tools by implementing BaseTool and registering in the configuration

Key files:

• server.py - Main MCP server with registry integration
• registry/ - Tool registry and loader
• tools/ - Tool implementations organized by category

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black src/

# Lint
ruff check src/

If you like this project, please give it a star ⭐

For questions, feedback, or support, reach out to:

Artem KK | MIT LICENSE