Tavily MCP Server

A powerful Model Context Protocol (MCP) server that integrates Tavily's AI-powered search and content extraction capabilities directly into your AI applications. This server enables you to perform sophisticated web searches with advanced filtering options and extract clean, structured content from any URL, making it ideal for research, content aggregation, and real-time information retrieval tasks.

Overview

The Tavily MCP Server provides seamless access to Tavily's search and extraction APIs through the Model Context Protocol. Whether you need to search the web with AI-enhanced relevance ranking, filter results by date ranges and domains, or extract clean content from web pages in markdown or plain text format, this server delivers enterprise-grade search capabilities with fine-grained control over results.

Features

AI-Powered Search

Advanced search algorithms optimized for AI and LLM applications
Intelligent result ranking and relevance scoring
Support for general, news, and finance-specific searches
Built-in LLM-generated answers to search queries
Automatic parameter configuration based on query intent

Comprehensive Filtering

Filter results by publication date ranges (day, week, month, year)
Custom date range filtering with start and end dates
Country-specific result boosting for localized searches
Domain inclusion and exclusion lists for precise control
Configurable search depth (basic or advanced) for optimal cost-performance balance

Rich Content Extraction

Clean, formatted content extraction from any URL
Support for markdown and plain text output formats
Advanced extraction mode with table and embedded content parsing
Batch extraction from multiple URLs simultaneously
Image and favicon extraction capabilities

Result Customization

Control the number of results returned (up to 20)
Include or exclude images in search results
Add image descriptions for enhanced context
Include raw HTML content when needed
Adjust content chunks per source for advanced searches

Tools

tavily_search

Perform an AI-powered web search using Tavily with extensive customization options.

Required Parameters:

query (string): The search query to execute

Optional Parameters:

topic (string): Category of search - choose from general, news, or finance. Default is general for broad searches, use news for real-time updates, or finance for financial market data
search_depth (string): Choose between basic (1 credit) or advanced (2 credits) for more comprehensive and relevant results
max_results (integer): Maximum number of search results to return, ranging from 0 to 20
include_answer (boolean): Include an LLM-generated direct answer to your query
include_images (boolean): Include relevant image search results
include_image_descriptions (boolean): Add AI-generated descriptions to images (requires include_images to be enabled)
include_raw_content (boolean): Include the cleaned HTML content of each result
include_favicon (boolean): Include favicon URLs for each result domain
time_range (string): Filter results by publication date - options include day, week, month, year (or abbreviated as d, w, m, y)
start_date (string): Filter results from this date onwards in YYYY-MM-DD format
end_date (string): Filter results up to this date in YYYY-MM-DD format
days (integer): For news topic searches, specify how many days back to search (minimum 1)
country (string): Boost results from a specific country (available for general topic only) - supports over 150 countries from Afghanistan to Zimbabwe
include_domains (array of strings): List of up to 300 domains to specifically include in results
exclude_domains (array of strings): List of up to 150 domains to exclude from results
chunks_per_source (integer): For advanced searches, specify 1-3 content chunks to extract per source
auto_parameters (boolean): Enable automatic parameter configuration based on query intent analysis

Use Cases:

Research and fact-checking with specific date ranges
News monitoring with real-time updates
Financial market research with finance-specific searches
Location-specific searches with country boosting
Domain-filtered searches for trusted sources
Multi-modal searches combining text and images

tavily_extract

Extract and parse clean content from one or more URLs, returning structured data ready for AI processing.

Required Parameters:

urls (array of strings): List of URLs to extract content from (must be valid URI format)

Optional Parameters:

format (string): Output format - choose markdown for structured formatting or text for plain text
extract_depth (string): Choose between basic (1 credit per 5 URLs) or advanced (2 credits per 5 URLs) for enhanced extraction including tables and embedded content
timeout (number): Maximum wait time in seconds for extraction, ranging from 1 to 60 seconds
include_images (boolean): Include a list of all images extracted from the URLs
include_favicon (boolean): Include favicon URLs for each extracted page

Use Cases:

Content aggregation from multiple sources
Article parsing for summarization
Data extraction from web pages
Converting web content to markdown for documentation
Batch processing of URLs for analysis
Image cataloging from web pages

Resource Templates

search

URI Pattern: tavily://search/{query}

Access cached or stored search results for a specific query. This resource template provides a standardized way to reference and retrieve search results, enabling efficient re-access of previously executed searches without redundant API calls.

extracted-content

URI Pattern: tavily://extracted-content/{url}

Access previously extracted and parsed content from a specific URL. This resource template allows you to reference extracted content in a consistent manner, making it easy to manage and retrieve cleaned web content across your application.

Advanced Capabilities

Intelligent Search Depth

The server offers two search depth options to balance between speed, cost, and result quality. Basic search provides fast, cost-effective results suitable for most queries, while advanced search employs more sophisticated algorithms and returns highly relevant results with additional context, ideal for complex research tasks or when precision is paramount.

Multi-Topic Search Support

Different search topics are optimized for different use cases. General search casts a wide net for comprehensive results, news search provides real-time updates with date-range filtering and recency prioritization, and finance search delivers specialized results from financial data sources, market analysis, and economic news outlets.

Flexible Date Filtering

The server supports multiple approaches to date filtering. Use pre-configured time ranges for quick filtering by day, week, month, or year. For precise control, specify exact start and end dates. When using the news topic, the days parameter provides a convenient way to search recent news within a specific timeframe.

Domain Control

Maintain full control over which sources appear in your results. The include_domains parameter allows you to create allowlists of trusted sources (up to 300 domains), while exclude_domains helps filter out unwanted sources (up to 150 domains). This is particularly useful for academic research, brand monitoring, or when you need to focus on specific publication types.

Content Chunking

For advanced searches, the chunks_per_source parameter enables extraction of multiple relevant segments from each source. This is especially valuable when dealing with long-form content, as it ensures you capture all relevant information from a single page rather than just a single excerpt.

Best Practices

When using the Tavily MCP Server, consider the search depth that matches your needs - basic for quick lookups and advanced for comprehensive research. Leverage domain filtering to improve result quality and reduce noise. For time-sensitive queries, combine the news topic with appropriate date filters. When extracting content, batch multiple URLs together for efficiency, and choose markdown format when you need to preserve document structure for further processing.

The auto_parameters feature is particularly powerful for dynamic applications where query intent varies - it analyzes your query and automatically optimizes search parameters for the best results.