Skip to content
MCP server

FATSEARCH

Turns web pages and PDFs into cached Markdown outlines read one section at a time

Runs locallyStatus and history
Pricing
Open source
Platforms
CLI, macOS, Linux
Runs
Locally
Version
0.2.5
Listing updated
2 days ago

About FATSEARCH

FATSEARCH is an open-source MCP server for web search and scraping. It runs locally on your machine and is started by your AI client.

FATSEARCH goes after the pages a built-in fetch tool reads badly: JavaScript shells, PDFs and pages behind captchas. It tries plain HTTP first and only climbs to your own paid providers, cheapest first, when that fails, tracking cost on every call. A dry run prints the planned route and an upper cost estimate before anything is spent.

Instead of dumping a whole page, it returns a short outline with section ids. In the README's Wikipedia example the page was 40.9k tokens and the agent used 1,384 tokens in total against 112,517 for a Firecrawl copy. Menus, banners and footers are removed while prices, dates and numbers are kept. Results are stored on disk as Markdown, and filters such as an exact phrase search return matching lines.

There is no hosted account: you bring a provider key, set through a one-time local form on 127.0.0.1 so the key never enters the chat or logs. The README reports 35 of 38 hard pages answered within 60 seconds, against 24 for Firecrawl and 19 for Tavily in its own benchmark, and notes the others are faster on pages they can read.

Install

Run in a terminal

claude mcp add fats -- npx -y fatsearch

Add to claude_desktop_config.json

{
  "mcpServers": {
    "fats": {
      "command": "npx",
      "args": [
        "-y",
        "fatsearch"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "fats": {
      "command": "npx",
      "args": [
        "-y",
        "fatsearch"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "fats": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "fatsearch"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "fats": {
      "command": "npx",
      "args": [
        "-y",
        "fatsearch"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "fats": {
      "command": "npx",
      "args": [
        "-y",
        "fatsearch"
      ]
    }
  }
}

Configuration

  • BRIGHTDATA_UNLOCKER_ZONEBright Data Web Unlocker zone name (optional; fats setup bd saves the account's zone)

What it does

Section-by-section reading
Returns an outline with ids so the agent opens only what it needs.
Escalating fetch
Starts with free plain HTTP and moves to your providers when blocked.
Keys stay out of chat
A local one-time form saves the key with file mode 0600.
Cost tracking
`fats stats` shows tokens, time and spend per page.
Local Markdown cache
Pages are kept on disk and searchable by phrase.
Single binary
Ships through cargo, npm and release downloads.

Good for

  • Read a JavaScript-heavy page an agent's fetch cannot
  • Extract one section of a long PDF or Wikipedia article
  • Cut the token cost of web research
  • Check what a fetch will cost before running it

Pricing

Open source

Free tool; provider fetch fees are paid to your own provider

Prices change. Check fatsearch.com before you buy.

FATSEARCH FAQ

Does it need an account?

No account with FATSEARCH itself; you provide your own provider key, for example Bright Data or Decodo.

Does it track me?

The site says there is no personal telemetry.

How is the key entered safely?

Through a one-time form on 127.0.0.1 behind a random token that expires after 30 minutes.

Which licence?

The README links an MIT licence file.

Is FATSEARCH free?

Yes. FATSEARCH is open source, so you can use and self-host it for free.

How do I add FATSEARCH to Claude Code?

Run this command in your terminal: claude mcp add fats -- npx -y fatsearch

What are the best alternatives to FATSEARCH?

The closest alternatives to FATSEARCH are Scrapling MCP Server, Firecrawl MCP Server, SearXNG Search, NeuralVerge and AI-Powered directory.

Alternatives to FATSEARCH

  1. MCP server for AI clients to scrape sites with HTTP and stealth browser sessions

    Runs locally
    • 17 tools
    • Open source
    • 104K GitHub stars
    • Web search and scraping
  2. MCP server that lets AI agents search, scrape and interact with web pages

    • 3 tools
    • Remote
    • Freemium
    • 7.8K GitHub stars
    • Web search and scraping
  3. SearXNG Search

    MCP server

    MCP server for web search and page reading through a SearXNG instance

    Runs locally
    • 3 tools
    • Open source
    • 1.4K GitHub stars
    • Web search and scraping
  4. NeuralVerge

    MCP server

    Structured, source-tagged business and people data from 300+ sources for agents

    • 200 tools
    • Remote
    • Free trial
    • 1 GitHub star
    • Web search and scraping
  5. Search AI tools, agents and MCP servers and get install configs from inside your AI assistant.

    • 4 tools
    • Remote
    • Free
    • Developer tools
  6. Hosted MCP server that returns any Shopify store's products, prices and stock

    Not checked yet
    • 4 tools
    • Remote
    • Paid
    • E-commerce