Tavily Search MCP Server - Fast AI Web Search for Agents

Search the web through AI assistants like Claude, ChatGPT, and OpenClaw. Get ranked results, AI answers, images, and full page content powered by Tavily's AI search engine.

No Tavily API key needed. Just connect and search.

Fast. Results in 2-3 seconds per tool call. No queue, no waiting.

1 tool with 16 parameters: web search with 4 depth levels, AI-generated answers, date filtering, domain restrictions, country boosting, news mode, full page content extraction, and image results.

Stateful MCP server running in Apify Standby mode. Persistent session with no cold starts.

MCP endpoint:

https://clearpath--tavily-search-mcp.apify.actor/mcp

 

Tavily Suite ➤	Search API	Search MCP Server

Tool

1 tool available to connected AI agents:

• search
  • Search the web using Tavily's AI search engine. Returns ranked results with relevance scores, snippets, and optional AI-generated answers, images, and full page content.
  • Inputs:
    • query (string, required): Search query text
    • search_depth (string, optional): "ultra-fast", "fast", "basic", or "advanced" (default "basic")
    • max_results (number, optional): 1-20 (default 5)
    • topic (string, optional): "general" or "news" (default "general")
    • include_answer (string, optional): "false", "basic", or "advanced" (default "false")
    • time_range (string, optional): "day", "week", "month", or "year"
    • start_date (string, optional): Results after this date (YYYY-MM-DD)
    • end_date (string, optional): Results before this date (YYYY-MM-DD)
    • exact_match (boolean, optional): Exact phrase match only (default false)
    • include_domains (array, optional): Only search these domains
    • exclude_domains (array, optional): Exclude these domains
    • country (string, optional): Boost results from this country
    • include_raw_content (string, optional): "false", "markdown", or "text" (default "false")
    • include_images (boolean, optional): Include image results (default false)
    • include_image_descriptions (boolean, optional): AI descriptions for images (default false)
    • chunks_per_source (number, optional): 1-3, advanced depth only (default 3)
  • ~$0.004 per credit (basic = 1 credit, advanced = 2 credits)

Output Example

{
    "query": "best AI frameworks 2026",
    "answer": "The most widely adopted AI frameworks in 2026 span deep-learning libraries, traditional machine-learning toolkits, and specialized agent-orchestration platforms: PyTorch and TensorFlow (with its high-level Keras API) remain the backbone for building and scaling neural-network models, while Scikit-learn...",
    "follow_up_questions": null,
    "images": [
        {
            "url": "https://scalevista.com/wp-content/uploads/2025/12/Best-AI-Frameworks-for-Developers.webp",
            "description": "A list of the top 10 AI frameworks for developers in January 2026, featuring tools like TensorFlow, PyTorch, Keras, and others."
        }
        // ... 4 more images
    ],
    "results": [
        {
            "url": "https://www.kellton.com/kellton-tech-blog/ai-tech-stack-2026",
            "title": "AI Tech Stack 2026: Frameworks, MLOps & IDEs Guide - Kellton",
            "content": "Integrated Development Environments specialized for AI development accelerate coding, debugging, and experimentation. Top IDEs for AI development in 2026...",
            "score": 0.9486564,
            "raw_content": "# AI Tech Stack 2026: Frameworks, MLOps, and IDEs reshaping AI Development\n\n[AI/ML](https://www.kellton.com/ai-ml)\n\nJanuary 15, 2026...",
            "favicon": "https://www.kellton.com/themes/custom/kellton_theme/favicon.ico"
        }
        // ... 4 more results
    ]
}

Quick Start

Basic search

"Search for the best restaurants in Berlin"

News search

"Search for latest AI news from the past week"

Deep search with AI answer

"Search for Python async best practices with an advanced AI answer"

Domain-restricted search

"Search reddit.com and github.com for FastAPI authentication examples"

How It Works

1. Agent connects to the MCP server endpoint
2. Searches using the search tool with any combination of parameters
3. Gets results structured for easy consumption by the AI agent

No Tavily API key needed. The server handles everything.

---

Pricing - Pay Per Event (PPE)

Search Depth	Credits	Cost
Ultra-fast	1	~$0.004
Fast	1	~$0.004
Basic	1	~$0.004
Advanced	2	~$0.008

AI-generated answers (basic or advanced) may use additional credits.

---

Configuration

No configuration required. The server works out of the box.

Replace YOUR_APIFY_TOKEN in all examples below with your Apify API token.

Usage with ChatGPT

ChatGPT supports MCP servers natively (Plus, Pro, Team, Enterprise plans).

1. Open Settings > Connectors > Add connector
2. Set Name to Reddit
3. Set MCP Server URL to:

https://clearpath--tavily-search-mcp.apify.actor/mcp

4. Set Authentication to API Key and set the header:
   • Header name: Authorization
   • Value: Bearer YOUR_APIFY_TOKEN
5. Click Save
6. Start a new chat, click + near the composer, select More, choose Reddit

Usage with Claude Desktop

Add to your claude_desktop_config.json:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
    "mcpServers": {
        "tavily-search": {
            "type": "http",
            "url": "https://clearpath--tavily-search-mcp.apify.actor/mcp",
            "headers": {
                "Authorization": "Bearer YOUR_APIFY_TOKEN"
            }
        }
    }
}

Restart Claude Desktop after saving.

Usage with Claude Code

claude mcp add --transport http --scope user tavily-search \
  --header "Authorization: Bearer YOUR_APIFY_TOKEN" \
  https://clearpath--tavily-search-mcp.apify.actor/mcp

Or add to your project's .mcp.json:

{
    "mcpServers": {
        "tavily-search": {
            "type": "http",
            "url": "https://clearpath--tavily-search-mcp.apify.actor/mcp",
            "headers": {
                "Authorization": "Bearer ${APIFY_TOKEN}"
            }
        }
    }
}

Usage with OpenClaw

Add to ~/.openclaw/openclaw.json:

{
  "provider": {
    "mcpServers": {
      "tavily-search": {
        "url": "https://clearpath--tavily-search-mcp.apify.actor/mcp",
        "headers": {
          "Authorization": "Bearer YOUR_APIFY_TOKEN"
        }
      }
    }
  }
}

Then message your OpenClaw bot on Telegram, WhatsApp, or Discord:

"Search for the latest news about AI regulation"

"Find me recent articles about climate change from the past month"

Other MCP Clients

Any client supporting the Streamable HTTP transport works. Use the MCP endpoint URL with your Apify API token as a Bearer auth header.

---

FAQ

Do I need a Tavily API key? No. No Tavily API key or Tavily account required. The server handles everything.

What's the difference between search depths? Ultra-fast and fast prioritize speed. Basic balances speed and relevance. Advanced provides the best relevance with multi-source content chunks.

What countries are supported? 195 countries. The country parameter boosts results from the specified country (general topic only).

Can I get full page content? Yes. Set include_raw_content to "markdown" or "text" to get the full parsed content of each result page.

How is this different from the Tavily Search API actor? The API actor runs as a batch job and stores results in a dataset. This MCP server provides real-time search as tool calls for AI agents.

Can I use this without an AI assistant? Yes. Any MCP client works. You can also call tools programmatically via any MCP SDK.

---

Support

• Email: max@mapa.slmail.me
• Bugs: Issues tab

---

Search the web through AI agents. Get ranked results, AI answers, and full page content powered by Tavily.