Zyte Fetch Page Content MCP Server

A Model Context Protocol (MCP) server that extracts clean, LLM-friendly content from any webpage using the Zyte API. Perfect for AI assistants that need to read and understand web content.

Features

🧹 Clean Content Extraction - Get main page content
🌐 Browser Rendering - Full JavaScript support for dynamic pages
⚡ Configurable Extraction - Choose between speed (HTTP) or quality (browser rendering)
🔒 Secure - API keys stored in Docker secrets, runs as non-root user
🐳 Docker-based - Easy deployment via Docker MCP Toolkit

Tools Available

The server provides three tools, each using a different extraction method:

Tool	Description
`fetch_page_content_from_http_response_body`	Extract from HTTP response (fastest, cheapest)
`fetch_page_content_from_browser_html`	Extract from browser with visual features (best quality, default)
`fetch_page_content_from_browser_html_only`	Extract from browser HTML only (better for JS-heavy pages)

Extraction Methods Explained

Method	Speed	Cost	Use Case	Returns
HTTP Response Body	Fastest	Cheapest	Simple static pages, quick previews	`pageContent` only
Browser HTML	Medium	Medium	Best quality, visual features included	`pageContent` + `browserHtml`
Browser HTML Only	Slower	Higher	JavaScript-heavy pages, SPAs	`pageContent` + `browserHtml`

Prerequisites

Docker Desktop with MCP Toolkit enabled
Docker MCP CLI plugin (docker mcp command)
Zyte API key (free tier available)

Installation

Step 1: Clone the Repository

git clone https://github.com/YOUR_USERNAME/zyte-fetch-page-content-mcp-server.git
cd zyte-fetch-page-content-mcp-server

Step 2: Build the Docker Image

docker build -t zyte-fetch-page-content-mcp-server .

Step 3: Set Your Zyte API Key

docker mcp secret set ZYTE_API_KEY="your-zyte-api-key-here"

Get your API key from Zyte Dashboard.

Step 4: Import the Catalog

# Create catalogs directory if it doesn't exist
mkdir -p ~/.docker/mcp/catalogs

# Copy catalog file
cp custom.yaml ~/.docker/mcp/catalogs/custom.yaml

# Import the catalog
docker mcp catalog import ~/.docker/mcp/catalogs/custom.yaml

Step 5: Add the Server

docker mcp server add zyte-fetch-page-content

Step 6: Verify Installation

# Check server is listed
docker mcp server ls

# Should show:
# zyte-fetch-page-content    -    ✓ done    -    Fetches page content...

# Check tools are available
docker mcp tools ls | grep fetch

Step 7: Configure LLM Client (e.g., Claude Desktop)

Edit your Claude Desktop config file:

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

Ensure it contains:

{
  "mcpServers": {
    "MCP_DOCKER": {
      "command": "docker",
      "args": ["mcp", "gateway", "run"]
    }
  }
}

Step 8: Restart Claude Desktop

Quit Claude Desktop completely (Cmd+Q / Ctrl+Q) and reopen it.

Usage

Once installed, you can ask Claude to use the tools:

Basic Usage (Best Quality)

Use fetch_page_content_from_browser_html to get content from https://example.com

Fast Extraction (Static Pages)

Use fetch_page_content_from_http_response_body to quickly get content from https://example.com

JavaScript-Heavy Pages

Use fetch_page_content_from_browser_html_only to extract content from https://example.com

Tool Parameters

All three tools accept a single parameter:

url (required): The URL to fetch content from

Example:

fetch_page_content_from_browser_html(url="https://example.com")

Response Format

All tools return raw JSON text from the Zyte API.

HTTP Response Body Response

{
  "url": "https://example.com",
  "statusCode": 200,
  "pageContent": {
    "itemMain": "Main article content as clean text...",
    "itemSecondary": "Sidebar and navigation content..."
  }
}

Browser HTML Response (with visual features)

{
  "url": "https://example.com",
  "statusCode": 200,
  "browserHtml": "<!DOCTYPE html><html>...</html>",
  "pageContent": {
    "itemMain": "Main article content as clean text...",
    "itemSecondary": "Sidebar and navigation content..."
  }
}

Browser HTML Only Response

{
  "url": "https://example.com",
  "statusCode": 200,
  "browserHtml": "<!DOCTYPE html><html>...</html>",
  "pageContent": {
    "itemMain": "Main article content as clean text...",
    "itemSecondary": "Sidebar and navigation content..."
  }
}

Testing

Test the Docker Image Directly

# Test that the server starts
echo '{"jsonrpc":"2.0","method":"tools/list","id":1}' | docker run --rm -i -e ZYTE_API_KEY="your-key" zyte-fetch-page-content-mcp-server

Test via Docker MCP

# Test HTTP response body extraction
docker mcp tools call fetch_page_content_from_http_response_body url="https://example.com"

# Test browser HTML extraction (best quality)
docker mcp tools call fetch_page_content_from_browser_html url="https://example.com"

# Test browser HTML only extraction
docker mcp tools call fetch_page_content_from_browser_html_only url="https://example.com"

Troubleshooting

Server Not Found in Catalog

# Reimport the catalog
docker mcp catalog rm custom
docker mcp catalog import ~/.docker/mcp/catalogs/custom.yaml
docker mcp server add zyte-fetch-page-content

or add this to ~/.docker/mcp/registry.yaml

registry:
  zyte-fetch-page-content:
    ref: ""

Tools Not Appearing in Claude

Verify server is enabled: docker mcp server ls
Check tools exist: docker mcp tools ls | grep fetch
Restart Claude Desktop completely

API Key Issues

# Verify secret is set
docker mcp secret ls | grep ZYTE

# Reset the secret
docker mcp secret set ZYTE_API_KEY="your-key"

Configuration Error Response

If you see this response:

{"error": "ZYTE_API_KEY not configured", "type": "config_error"}

The API key is not set correctly. Run:

docker mcp secret set ZYTE_API_KEY="your-actual-key"

Timeout Errors

The default timeout is 120 seconds. If you experience timeouts:

Use fetch_page_content_from_http_response_body for faster results on simple pages
Consider that complex pages with heavy JavaScript may take longer to render

How It Works

Request Flow

Claude Desktop
      ↓
MCP Gateway (Docker)
      ↓
Zyte MCP Server (this project)
      ↓
Zyte API (api.zyte.com)
      ↓
Returns JSON with pageContent

Implementation Details

Each tool makes an HTTPS POST request to https://api.zyte.com/v1/extract with:

Authentication: HTTP Basic Auth using your Zyte API key
Timeout: 120 seconds
Headers: Content-Type: application/json, Accept-Encoding: gzip
Payload: URL and extraction options

The tools differ only in their pageContentOptions.extractFrom value and whether browserHtml: true is included.

Development

Local Development

# Install dependencies
pip install mcp[cli] httpx

# Set API key
export ZYTE_API_KEY="your-key"

# Run server
python zyte_fetch_page_content_server.py

Test MCP Protocol Locally

# List available tools
echo '{"jsonrpc":"2.0","method":"tools/list","id":1}' | python zyte_fetch_page_content_server.py

Rebuild After Changes

docker build -t zyte-fetch-page-content-mcp-server .
docker mcp server rm zyte-fetch-page-content
docker mcp server add zyte-fetch-page-content
# Restart Claude Desktop

Security

API keys are stored in Docker Desktop secrets (never in code)
Container runs as non-root user
No sensitive data is logged
All network requests use HTTPS
HTTP Basic Auth for API authentication

Performance Tips

For static HTML pages: Use fetch_page_content_from_http_response_body (fastest)
For best quality extraction: Use fetch_page_content_from_browser_html (includes visual features)
For JavaScript-heavy SPAs: Use fetch_page_content_from_browser_html_only
Cost optimization: HTTP response body extraction is the cheapest option

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
Dockerfile		Dockerfile
LICENSE		LICENSE
README.md		README.md
custom.yaml		custom.yaml
requirements.txt		requirements.txt
zyte_fetch_page_content_server.py		zyte_fetch_page_content_server.py

Folders and files

Latest commit

History

Repository files navigation

Zyte Fetch Page Content MCP Server

Features

Tools Available

Extraction Methods Explained

Prerequisites

Installation

Step 1: Clone the Repository

Step 2: Build the Docker Image

Step 3: Set Your Zyte API Key

Step 4: Import the Catalog

Step 5: Add the Server

Step 6: Verify Installation

Step 7: Configure LLM Client (e.g., Claude Desktop)

Step 8: Restart Claude Desktop

Usage

Basic Usage (Best Quality)

Fast Extraction (Static Pages)

JavaScript-Heavy Pages

Tool Parameters

Response Format

HTTP Response Body Response

Browser HTML Response (with visual features)

Browser HTML Only Response

Testing

Test the Docker Image Directly

Test via Docker MCP

Troubleshooting

Server Not Found in Catalog

Tools Not Appearing in Claude

API Key Issues

Configuration Error Response

Timeout Errors

How It Works

Request Flow

Implementation Details

Development

Local Development

Test MCP Protocol Locally

Rebuild After Changes

Security

Performance Tips

Useful Links

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages