Skip to content

MCP

Microlink MCP exposes Microlink API as MCP tools for assistants like Claude, Codex, Cursor, and VS Code.

Install

No global install is required. Run with npx (or pnpm dlx if you use pnpm):
npx -y @microlink/mcp
pnpm dlx @microlink/mcp

Add it to your MCP client

Use this mcpServers entry:
{
  "mcpServers": {
    "microlink": {
      "command": "npx",
      "args": ["-y", "@microlink/mcp"],
      "env": {
        // Optional. Free 50 reqs/day rate limit.
        "MICROLINK_API_KEY": "YOUR_MICROLINK_API_KEY"
      }
    }
  }
}
For higher/unlimited usage, get an API key at .
If no API key is provided, requests go to the free endpoint with a 50 requests/day limit.

Usage

Once you've added the MCP to your favorite LLM client, try asking it to take a screenshot of a specific webpage, generate a PDF, or fetch metadata. Just use plain language to describe what you want — the agent will know which tool to call and with what parameters.

Tools

  • microlink_extract: General extractor. Metadata + custom data scraping + optional screenshot/pdf/video/audio/insights/palette in one call.
  • microlink_screenshot: Capture screenshots (fullPage, element, overlay, type, etc.).
  • microlink_pdf: Generate PDFs (format, margin, scale, pageRanges, etc.).
  • microlink_video: Extract playable video source (data.video).
  • microlink_audio: Extract playable audio source (data.audio).
  • microlink_insights: Lighthouse + technology detection.
  • microlink_meta: Normalized metadata only (title, description, image, logo, etc.).
  • microlink_palette: Extract color palette from detected images.
  • microlink_markdown: Convert URL content to Markdown (microlink.data.markdown).
  • microlink_text: Convert URL content to plain text (microlink.data.text).
Notes:
  • screenshot, pdf, and insights accept true or { ...options }.
  • Empty objects like {} are treated as true.
  • Boolean strings ("true", "false") and JSON-stringified objects are normalized for compatibility with some MCP clients.

Response format

  • All tools return structuredContent with:
    • endpoint, requestUrl, finalUrl, statusCode, responseHeaders, microlink
  • For markdown/text specifically:
    • Markdown content is in microlink.data.markdown
    • Plain text content is in microlink.data.text

Prompt examples

Extract metadata + screenshot in one request

"Get a full page screenshot of microlink.io and include the metadata"
{
  "tool": "microlink_extract",
  "arguments": {
    "url": "https://microlink.io",
    "screenshot": {
      "fullPage": true
    },
    "meta": true
  }
}

Screenshot with overlay

"Get a screenshot of microlink.io with a dark browser overlay"
{
  "tool": "microlink_screenshot",
  "arguments": {
    "url": "https://microlink.io",
    "screenshot": {
      "overlay": { "browser": "dark" },
      "type": "png"
    }
  }
}

Generate a PDF

"Download a pdf of example.com"
{
  "tool": "microlink_pdf",
  "arguments": {
    "url": "https://example.com",
    "pdf": {
      "format": "A4",
      "margin": "0.35cm"
    }
  }
}

Get page insights

"Check the performance of microlink.io"
{
  "tool": "microlink_insights",
  "arguments": {
    "url": "https://microlink.io",
    "insights": {
      "lighthouse": { "preset": "perf" },
      "technologies": true
    }
  }
}

Convert a page to markdown

"Read as markdown and give the key points of the article"
{
  "tool": "microlink_markdown",
  "arguments": {
    "url": "https://microlink.io/blog/antibot-detection-at-scale"
  }
}