@leonbeckert/mcp-server-zenrows v0.1.0

ZenRows Fetch MCP Server

An Model Context Protocol (MCP) server that retrieves web pages through ZenRows and returns a clean, Markdown‑formatted snapshot—while automatically escalating from the cheapest request mode to more advanced (and expensive) bypass levels only when needed.

✨ Features

• Cost‑aware escalation – starts with the base ZenRows request and retries with Premium Proxy → Stealth (Premium + JS) → Stealth + Wait (2.5 s) until useful content is obtained.
• LLM‑ready output – delivers Markdown/text instead of raw HTML, ideal for downstream summarisation or analysis.
• Optional selector wait – pass a CSS selector to wait for (implies JS rendering) when working with SPAs or late‑loading content.
• Blocked‑page detection – heuristically recognises captchas/forbidden responses and escalates automatically.
• Cost tier metadata – the final tier used (basic, premium, stealth, or wait) is returned in additionalParams.meta.costTier.
• Zero‑config runtime – ship as an npx one‑liner, a slim Docker image, or run directly on Node 22+.

🛠️ Tool

zenrowsFetch

Fetches a web page through ZenRows and returns its Markdown rendition.

Output – Plain‑text (Markdown) content of the page. The enclosing MCP response includes additionalParams.meta.costTier.

🚀 Usage

Use zenrowsFetch whenever you need a lightweight, resilient webpage retriever that:

• Minimises cost by default but adapts when the target site blocks basic scraping.
• Produces clean, LLM‑friendly Markdown with no extra parsing steps.
• Works equally well for simple HTML pages and heavy JavaScript single‑page apps.

Typical scenarios include web‑content summarisation, ad‑hoc research pipelines, and programmatic ingestion of news/article sites.

⚙️ Configuration

Claude Desktop

npx

{
	"mcpServers": {
		"zenrows-fetch": {
			"command": "npx",
			"args": ["-y", "@leonbeckert/mcp-server-zenrows"],
			"env": {
				"ZENROWS_API_KEY": "zenrw_…"
			}
		}
	}
}

docker

{
	"mcpServers": {
		"zenrows-fetch": {
			"command": "docker",
			"args": [
				"run",
				"--rm",
				"-i",
				"-e",
				"ZENROWS_API_KEY=zenrw_…",
				"leonbeckert/mcp-server-zenrows"
			]
		}
	}
}

🧑‍💻 Local Install (No Docker, No NPM Publish)

This lets you run the server locally while using npx to launch it from tools like Claude Desktop.

🧱 Step-by-Step

1. Clone or Set Up the Code Locally

mkdir -p ~/code/zenrows-mcp
cd ~/code/zenrows-mcp
# Copy the relevant files here (index.ts, package.json, etc.)

2. Install Dependencies and Build

npm install       # installs dependencies: zenrows, MCP SDK, TypeScript types
npm run build     # compiles TypeScript into dist/, and makes dist/index.js executable

3. Globally Link the Binary

npm link

This makes the mcp-server-zenrows binary globally accessible via npx.

4. Smoke Test

ZENROWS_API_KEY=zenrw_test npx mcp-server-zenrows

You should see:

ZenRows MCP server running on stdio

5. Update mcp.json / user.settings.json Locally

{
	"mcp": {
		"servers": {
			"zenrows-fetch": {
				"command": "npx",
				"args": ["-y", "mcp-server-zenrows"],
				"env": {
					"ZENROWS_API_KEY": "zenrw_…"
				}
			}
		}
	}
}

This will launch your local server via the symlink created by npm link.

🧪 One-click in VS Code

Install via NPX

Install via Docker

🏗️ Building the Docker Image

# Build minimal production image
docker build -t mcp/zenrows .

🪪 License

MIT License – see the LICENSE file for full details.