mcp-omnisearch

A Model Context Protocol (MCP) server that provides unified access to multiple search providers and AI tools. This server combines the capabilities of Tavily, Perplexity, Kagi, Jina AI, Brave, Exa AI, and Firecrawl to offer comprehensive search, AI responses, content processing, and enhancement features through a single interface.

Features

🔍 Search Tools

• Tavily Search: Optimized for factual information with strong citation support. Supports domain filtering through API parameters (include_domains/exclude_domains).

• Brave Search: Privacy-focused search with comprehensive operator support: site:, -site:, filetype:/ext:, intitle:, inurl:, inbody:, inpage:, lang:, loc:, before:, after:, +term, -term, and exact phrases ("phrase").

• Kagi Search: High-quality search with full operator support: site:, -site:, filetype:/ext:, intitle:, inurl:, inbody:, inpage:, lang:, loc:, before:, after:, +term, -term, and exact phrases ("phrase").

• Exa Search: AI-powered web search using neural and keyword search. Optimized for AI applications with semantic understanding, content extraction, and research capabilities.

• GitHub Search: Comprehensive code search across public GitHub repositories with three specialized tools:

  • Code Search: Find code examples, function definitions, and files using advanced syntax (filename:, path:, repo:, user:, language:, in:file)

  • Repository Search: Discover repositories with sorting by stars, forks, or recent updates

  • User Search: Find GitHub users and organizations

🎯 Search Operators

MCP Omnisearch provides powerful search capabilities through operators and parameters:

Search Operator Reference

Brave & Kagi Operators (use in query string):

• Domain: site:example.com, -site:example.com

• File type: filetype:pdf or ext:pdf

• Location: intitle:term, inurl:term, inbody:term, inpage:term

• Language: lang:en (ISO 639-1 codes)

• Country: loc:us (ISO 3166-1 codes)

• Date: before:2024, after:2024-01-01

• Exact: "exact phrase"

• Include/Exclude: +required, -excluded

Tavily (API parameters only):

• Domain filtering: include_domains, exclude_domains

Example Usage

Provider Capabilities

• Brave Search: Full native operator support in query string

• Kagi Search: Complete operator support in query string

• Tavily Search: Domain filtering through API parameters

• Exa Search: Domain filtering through API parameters, semantic search with neural understanding

• GitHub Search: Advanced code search syntax with qualifiers:

  • filename:remote.ts - Search for specific files

  • path:src/lib - Search within specific directories

  • repo:user/repo - Search within specific repositories

  • user:username - Search within a user's repositories

  • language:typescript - Filter by programming language

  • in:file "export function" - Search for text within files

🤖 AI Response Tools

• Perplexity AI: Advanced response generation combining real-time web search with GPT-4 Omni and Claude 3

• Kagi FastGPT: Quick AI-generated answers with citations (900ms typical response time)

• Exa Answer: Get direct AI-generated answers to questions using Exa Answer API

📄 Content Processing Tools

• Jina AI Reader: Clean content extraction with image captioning and PDF support

• Kagi Universal Summarizer: Content summarization for pages, videos, and podcasts

• Tavily Extract: Extract raw content from single or multiple web pages with configurable extraction depth ('basic' or 'advanced'). Returns both combined content and individual URL content, with metadata including word count and extraction statistics

• Firecrawl Scrape: Extract clean, LLM-ready data from single URLs with enhanced formatting options

• Firecrawl Crawl: Deep crawling of all accessible subpages on a website with configurable depth limits

• Firecrawl Map: Fast URL collection from websites for comprehensive site mapping

• Firecrawl Extract: Structured data extraction with AI using natural language prompts

• Firecrawl Actions: Support for page interactions (clicking, scrolling, etc.) before extraction for dynamic content

• Exa Contents: Extract full content from Exa search result IDs

• Exa Similar: Find web pages semantically similar to a given URL using Exa

🔄 Enhancement Tools

• Kagi Enrichment API: Supplementary content from specialized indexes (Teclis, TinyGem)

• Jina AI Grounding: Real-time fact verification against web knowledge

Related MCP server: MCP2Tavily

Flexible API Key Requirements

MCP Omnisearch is designed to work with the API keys you have available. You don't need to have keys for all providers - the server will automatically detect which API keys are available and only enable those providers.

For example:

• If you only have a Tavily and Perplexity API key, only those providers will be available

• If you don't have a Kagi API key, Kagi-based services won't be available, but all other providers will work normally

• The server will log which providers are available based on the API keys you've configured

This flexibility makes it easy to get started with just one or two providers and add more as needed.

Configuration

This server requires configuration through your MCP client. Here are examples for different environments:

Cline Configuration

Add this to your Cline MCP settings:

Claude Desktop with WSL Configuration

For WSL environments, add this to your Claude Desktop configuration:

Environment Variables

The server uses API keys for each provider. You don't need keys for all providers - only the providers corresponding to your available API keys will be activated:

• TAVILY_API_KEY: For Tavily Search

• PERPLEXITY_API_KEY: For Perplexity AI

• KAGI_API_KEY: For Kagi services (FastGPT, Summarizer, Enrichment)

• JINA_AI_API_KEY: For Jina AI services (Reader, Grounding)

• BRAVE_API_KEY: For Brave Search

• GITHUB_API_KEY: For GitHub search services (Code, Repository, User search)

• EXA_API_KEY: For Exa AI services (Search, Answer, Contents, Similar)

• FIRECRAWL_API_KEY: For Firecrawl services (Scrape, Crawl, Map, Extract, Actions)

• FIRECRAWL_BASE_URL: For self-hosted Firecrawl instances (optional, defaults to Firecrawl cloud service)

You can start with just one or two API keys and add more later as needed. The server will log which providers are available on startup.

GitHub API Key Setup

To use GitHub search features, you'll need a GitHub personal access token with public repository access only for security:

1. Go to GitHub Settings: Navigate to GitHub Settings > Developer settings > Personal access tokens

2. Create a new token: Click "Generate new token" → "Generate new token (classic)"

3. Configure token settings:

   • Name: MCP Omnisearch - Public Search

   • Expiration: Choose your preferred expiration (90 days recommended)

   • Scopes: Leave all checkboxes UNCHECKED

     ⚠️ Important: Do not select any scopes. An empty scope token can only access public repositories and user profiles, which is exactly what we want for search functionality.

4. Generate and copy: Click "Generate token" and copy the token immediately

5. Add to environment: Set GITHUB_API_KEY=your_token_here

Security Notes:

• This token configuration ensures no access to private repositories

• Only public code search, repository discovery, and user profiles are accessible

• Rate limits: 5,000 requests/hour for code search, 10 requests/minute for code search specifically

• You can revoke the token anytime from GitHub settings if needed

Self-Hosted Firecrawl Configuration

If you're running a self-hosted instance of Firecrawl, you can configure MCP Omnisearch to use it by setting the FIRECRAWL_BASE_URL environment variable. This allows you to maintain complete control over your data processing pipeline.

Self-hosted Firecrawl setup:

1. Follow the Firecrawl self-hosting guide

2. Set up your Firecrawl instance (default runs on http://localhost:3002)

3. Configure MCP Omnisearch with your self-hosted URL:

Important notes:

• If FIRECRAWL_BASE_URL is not set, MCP Omnisearch will default to the Firecrawl cloud service

• Self-hosted instances support the same API endpoints (/v1/scrape, /v1/crawl, etc.)

• You'll still need a FIRECRAWL_API_KEY even for self-hosted instances

• Self-hosted Firecrawl provides enhanced security and customization options

API

The server implements MCP Tools organized by category:

Search Tools

search_tavily

Search the web using Tavily Search API. Best for factual queries requiring reliable sources and citations.

Parameters:

• query (string, required): Search query

Example:

search_brave

Privacy-focused web search with good coverage of technical topics.

Parameters:

• query (string, required): Search query

Example:

search_kagi

High-quality search results with minimal advertising influence. Best for finding authoritative sources and research materials.

Parameters:

• query (string, required): Search query

• language (string, optional): Language filter (e.g., "en")

• no_cache (boolean, optional): Bypass cache for fresh results

Example:

github_search

Search for code on GitHub using advanced syntax. This tool searches through file contents in public repositories and provides code snippets with metadata.

Parameters:

• query (string, required): Search query with GitHub search syntax

• limit (number, optional): Maximum number of results (1-50, default: 10)

Example:

Advanced query examples:

• "filename:config.json path:src" - Find config.json files in src directories

• "function fetchData language:typescript" - Find fetchData functions in TypeScript

• "repo:microsoft/vscode extension" - Search within specific repository

• "user:torvalds language:c" - Search user's repositories for C code

github_repository_search

Discover GitHub repositories with enhanced metadata including stars, forks, language, and last update information.

Parameters:

• query (string, required): Repository search query

• limit (number, optional): Maximum number of results (1-50, default: 10)

• sort (string, optional): Sort results by 'stars', 'forks', or 'updated'

Example:

github_user_search

Find GitHub users and organizations with profile information.

Parameters:

• query (string, required): User/organization search query

• limit (number, optional): Maximum number of results (1-50, default: 10)

Example:

exa_search

AI-powered web search using neural and keyword search. Automatically chooses between traditional keyword search and Exa's embeddings-based model to find the most relevant results for your query.

Parameters:

• query (string, required): Search query

• limit (number, optional): Maximum number of results (1-100, default: 10)

• include_domains (array, optional): Only include results from these domains

• exclude_domains (array, optional): Exclude results from these domains

Example:

AI Response Tools

ai_perplexity

AI-powered response generation with real-time web search integration.

Parameters:

• query (string, required): Question or topic for AI response

Example:

ai_kagi_fastgpt

Quick AI-generated answers with citations.

Parameters:

• query (string, required): Question for quick AI response

Example:

exa_answer

Get direct AI-generated answers to questions using Exa Answer API.

Parameters:

• query (string, required): Question for AI response

• include_domains (array, optional): Only include sources from these domains

• exclude_domains (array, optional): Exclude sources from these domains

Example:

Content Processing Tools

process_jina_reader

Convert URLs to clean, LLM-friendly text with image captioning.

Parameters:

• url (string, required): URL to process

Example:

process_kagi_summarizer

Summarize content from URLs.

Parameters:

• url (string, required): URL to summarize

Example:

process_tavily_extract

Extract raw content from web pages with Tavily Extract.

Parameters:

• url (string | string[], required): Single URL or array of URLs to extract content from

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced'

Example:

Response includes:

• Combined content from all URLs

• Individual raw content for each URL

• Metadata with word count, successful extractions, and any failed URLs

firecrawl_scrape_process

Extract clean, LLM-ready data from single URLs with enhanced formatting options.

Parameters:

• url (string | string[], required): Single URL or array of URLs to extract content from

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced'

Example:

Response includes:

• Clean, markdown-formatted content

• Metadata including title, word count, and extraction statistics

firecrawl_crawl_process

Deep crawling of all accessible subpages on a website with configurable depth limits.

Parameters:

• url (string | string[], required): Starting URL for crawling

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced' (controls crawl depth and limits)

Example:

Response includes:

• Combined content from all crawled pages

• Individual content for each page

• Metadata including title, word count, and crawl statistics

firecrawl_map_process

Fast URL collection from websites for comprehensive site mapping.

Parameters:

• url (string | string[], required): URL to map

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced' (controls map depth)

Example:

Response includes:

• List of all discovered URLs

• Metadata including site title and URL count

firecrawl_extract_process

Structured data extraction with AI using natural language prompts.

Parameters:

• url (string | string[], required): URL to extract structured data from

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced'

Example:

Response includes:

• Structured data extracted from the page

• Metadata including title, extraction statistics

firecrawl_actions_process

Support for page interactions (clicking, scrolling, etc.) before extraction for dynamic content.

Parameters:

• url (string | string[], required): URL to interact with and extract content from

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced' (controls complexity of interactions)

Example:

Response includes:

• Content extracted after performing interactions

• Description of actions performed

• Screenshot of the page (if available)

• Metadata including title and extraction statistics

exa_contents_process

Extract full content from Exa search result IDs.

Parameters:

• ids (string | string[], required): Exa search result ID(s) to extract content from

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced'

Example:

Response includes:

• Combined content from all result IDs

• Individual raw content for each ID

• Metadata with word count and extraction statistics

exa_similar_process

Find web pages semantically similar to a given URL using Exa.

Parameters:

• url (string, required): URL to find similar pages for

• extract_depth (string, optional): Extraction depth - 'basic' (default) or 'advanced'

Example:

Response includes:

• Combined content from all similar pages

• Similarity scores and metadata

• Individual content for each similar page

Enhancement Tools

enhance_kagi_enrichment

Get supplementary content from specialized indexes.

Parameters:

• query (string, required): Query for enrichment

Example:

enhance_jina_grounding

Verify statements against web knowledge.

Parameters:

• statement (string, required): Statement to verify

Example:

Docker Deployment

MCP Omnisearch supports containerized deployment using Docker with MCPO (Model Context Protocol Over HTTP) integration, enabling cloud deployment and OpenAPI access.

Quick Start with Docker

1. Using Docker Compose (Recommended):

2. Using Docker directly:

Container Environment Variables

Configure the container using environment variables for each provider:

• TAVILY_API_KEY: For Tavily Search

• PERPLEXITY_API_KEY: For Perplexity AI

• KAGI_API_KEY: For Kagi services (FastGPT, Summarizer, Enrichment)

• JINA_AI_API_KEY: For Jina AI services (Reader, Grounding)

• BRAVE_API_KEY: For Brave Search

• GITHUB_API_KEY: For GitHub search services

• EXA_API_KEY: For Exa AI services

• FIRECRAWL_API_KEY: For Firecrawl services

• FIRECRAWL_BASE_URL: For self-hosted Firecrawl instances (optional)

• PORT: Container port (defaults to 8000)

OpenAPI Access

Once deployed, the MCP server is accessible via OpenAPI at:

• Base URL: http://your-container-host:8000

• OpenAPI Endpoint: /omnisearch

• Compatible with: OpenWebUI and other tools expecting OpenAPI

Cloud Deployment

The containerized version can be deployed to any container platform that supports Docker:

• Cloud Run (Google Cloud)

• Container Instances (Azure)

• ECS/Fargate (AWS)

• Railway, Render, Fly.io

• Any Kubernetes cluster

Example deployment to a cloud platform:

Development

Setup

1. Clone the repository

2. Install dependencies:

3. Build the project:

4. Run in development mode:

Publishing

1. Update version in package.json

2. Build the project:

3. Publish to npm:

Troubleshooting

API Keys and Access

Each provider requires its own API key and may have different access requirements:

• Tavily: Requires an API key from their developer portal

• Perplexity: API access through their developer program

• Kagi: Some features limited to Business (Team) plan users

• Jina AI: API key required for all services

• Brave: API key from their developer portal

• GitHub: Personal access token with no scopes selected (public access only)

• Exa AI: API key from their dashboard at dashboard.exa.ai

• Firecrawl: API key required from their developer portal

Rate Limits

Each provider has its own rate limits. The server will handle rate limit errors gracefully and return appropriate error messages.

Contributing

Please read CONTRIBUTING.md before opening a PR. In short:

• Start by opening an issue to propose your change and align scope.

• Prefer small, focused PRs with a clear explanation (problem → approach → verification).

• Follow provider conventions: use src/common/http.ts (http_json) for HTTP, read keys from src/config/env.ts, respect timeouts, and surface errors via ProviderError.

License

MIT License - see the LICENSE file for details.

Acknowledgments

Built on:

• Model Context Protocol

• Tavily Search

• Perplexity AI

• Kagi Search

• Jina AI

• Brave Search

• Exa AI

• Firecrawl