Skip to content

The web browser your
AI agent never had

One config block gives Claude, Cursor, and any MCP client access to screenshots, PDFs, web scraping, markdown conversion, and metadata extraction. No API wrangling, no extra SDKs.

Read the docs
View on GitHub
mcp.json
{ "mcpServers": { "microlink": { "command": "npx", "args": ["-y", "@microlink/mcp"], "env": { "MICROLINK_API_KEY": "your-api-key" // Optional - Free tier 25 reqs/day } } } }
Paste this into your MCP client config file.
Claude Desktop
,
Cursor
,
Codex
, and every other MCP-compatible client gets access immediately.

Ask. Get. Done.

Use the Microlink API through natural language.
Tell your AI what you need. It screenshots, scrapes, converts, and extracts — straight from the live web. Type a request. Get a result.

Twenty tools. Zero boilerplate.

Everything Microlink can do, available to your AI through natural language. No HTTP clients, no API keys on day one, no parsing layers. Read the documentation to get started.

  • screenshot

    Ask your AI to screenshot any page and get back a CDN URL in seconds. Full-page, element crops, animated capture, device emulation, browser overlays, custom CSS/JS.

  • pdf

    Turn any URL into a downloadable PDF on demand. Your AI controls paper size, margins, orientation, scale, and page range — no headless browser setup required.

  • markdown

    Any webpage, clean Markdown output. 80% fewer tokens than raw HTML. Feed articles, docs, and research straight into your LLM context window.

  • extract

    The power tool. Scrape custom fields with CSS selectors and compose a screenshot, a PDF, and metadata in a single request. Structured JSON back.

  • metadata

    Fast, normalized metadata from any URL — title, description, author, date, and image. Plus a dedicated logo tool for pulling brand assets.

  • video & audio

    Extract a direct playable source from any media page. Video covers YouTube, Vimeo, TikTok, and Instagram; audio covers SoundCloud, Spotify, Mixcloud, and more.

  • search

    Google as structured data. Ranked results, knowledge graph, people-also-ask, and related searches — across news, images, videos, places, and shopping.

  • function

    Run your own JavaScript inside Microlink's server-side browser sandbox. Automate the clicks, scrolls, and scraping the API doesn't cover out of the box.

  • technologies

    Detect the full stack behind any site via Wappalyzer — frameworks, CDNs, analytics, and e-commerce platforms your competitors are running.

  • lighthouse

    Audit any URL for performance, accessibility, best practices, and SEO with Google Lighthouse — every failing rule listed, before launch not after.

  • collect

    Pull every link, image, video, audio source, or email address off a page in one call. Absolute, deduped arrays — ready to crawl or archive.

  • embed

    Get the oEmbed-ready iframe for YouTube, Tweets, CodePen, and hundreds more — the markup plus the scripts it needs, no manual wiring.

  • screenshot

    Ask your AI to screenshot any page and get back a CDN URL in seconds. Full-page, element crops, animated capture, device emulation, browser overlays, custom CSS/JS.

  • pdf

    Turn any URL into a downloadable PDF on demand. Your AI controls paper size, margins, orientation, scale, and page range — no headless browser setup required.

  • markdown

    Any webpage, clean Markdown output. 80% fewer tokens than raw HTML. Feed articles, docs, and research straight into your LLM context window.

  • extract

    The power tool. Scrape custom fields with CSS selectors and compose a screenshot, a PDF, and metadata in a single request. Structured JSON back.

  • metadata

    Fast, normalized metadata from any URL — title, description, author, date, and image. Plus a dedicated logo tool for pulling brand assets.

  • video & audio

    Extract a direct playable source from any media page. Video covers YouTube, Vimeo, TikTok, and Instagram; audio covers SoundCloud, Spotify, Mixcloud, and more.

  • search

    Google as structured data. Ranked results, knowledge graph, people-also-ask, and related searches — across news, images, videos, places, and shopping.

  • function

    Run your own JavaScript inside Microlink's server-side browser sandbox. Automate the clicks, scrolls, and scraping the API doesn't cover out of the box.

  • technologies

    Detect the full stack behind any site via Wappalyzer — frameworks, CDNs, analytics, and e-commerce platforms your competitors are running.

  • lighthouse

    Audit any URL for performance, accessibility, best practices, and SEO with Google Lighthouse — every failing rule listed, before launch not after.

  • collect

    Pull every link, image, video, audio source, or email address off a page in one call. Absolute, deduped arrays — ready to crawl or archive.

  • embed

    Get the oEmbed-ready iframe for YouTube, Tweets, CodePen, and hundreds more — the markup plus the scripts it needs, no manual wiring.

Unlock the web for your agents

Even behind bot detection, cookie walls, and ads.

screenshot
“Screenshot the Stripe pricing page with a dark browser overlay and share it with the team”
A framed, browser-chrome PNG hosted on CDN — ready to paste into Notion or Slack.
pdf
“Convert this documentation page to an A4 PDF and send it to the client”
A print-ready PDF with full CSS rendering, hosted on CDN and available immediately.
embed
“Give me the embeddable iframe for this YouTube video”
oEmbed-ready { html, scripts } — drop it straight into your page, no manual markup.
markdown
“Read this research paper and summarise the key findings in bullet points”
Clean Markdown — 80 % fewer tokens than raw HTML — fed straight into the LLM context.
html
“Get the raw HTML of this product page, scoped to the main article node”
The page HTML scoped to your selector — server-rendered by a real browser, ready to parse.
text
“Extract the plain text content of this article so I can count the words”
Raw readable text — no tags, no scripts, no noise — ready for analysis or summarisation.
metadata
“Fetch the title, OG image, and description for each of these 10 blog posts”
Normalized metadata for every URL — title, description, image, author, date, and logo.
logo
“Get the brand logo for stripe.com in a square shape”
The logo asset — url, type, and dimensions — with a square icon variant when you ask for one.
video
“Extract the playable video URL from this YouTube page so I can embed it”
A direct playable source URL — works with YouTube, Vimeo, TikTok, and hundreds more.
audio
“Extract the audio stream from this SoundCloud track so I can embed it”
A direct playable audio URL from SoundCloud — no API key, no scraping setup, no rate limits.
links
“Collect every link on this blog index page”
An array of absolute, deduped URLs — scope it to a CSS selector when you need a subset.
images
“Grab every image URL from this product gallery”
An array of absolute, deduped image URLs — scoped to a selector if you want just the gallery.
videos
“List every video source embedded on this page”
An array of absolute, deduped <video>/<source> URLs — the whole page, not just the primary one.
audios
“Collect every audio file linked on this podcast page”
An array of absolute, deduped <audio>/<source> URLs from across the page.
emails
“Find every email address on this company's contact page”
A deduped array of addresses — pulled from mailto: links and plain text alike.
technologies
“Detect what stack our competitor's site is running”
A Wappalyzer report — frameworks, CDNs, analytics, e-commerce platforms, and more.
lighthouse
“Audit our checkout page for performance and flag the biggest bottlenecks”
Lighthouse scores for performance, accessibility, best-practices, and SEO — with the failing rules.
search
“Search Google for the top MCP servers and give me structured results”
Ranked results — title, url, description — plus knowledge graph and related searches. Needs an API key.
function
“Run a script on this page to grab the price after the JS loads”
Your JavaScript executed in Microlink's server-side browser — the return value comes straight back.
extract
“Scrape every pricing plan name and monthly price from this SaaS page”
Structured JSON with the exact fields you asked for, extracted via CSS selectors.

What is Microlink MCP?

Microlink MCP is a Model Context Protocol server that gives AI assistants direct access to the Microlink API. Claude, Cursor, Windsurf, and any other MCP-compatible client can take screenshots, generate PDFs, scrape structured data, convert pages to markdown, and extract metadata — through natural language.
MCP is an open standard for connecting AI applications to external tools. Microlink MCP implements that standard for browser and web-data capabilities.

Do I need an API key?

No API key required to get started. The free tier covers 25 requests per day — enough to explore the toolset. The one exception is Search, which runs on paid plans and needs an API key.
Add your Microlink API key when you need Search, production volume, configurable TTL, custom headers, or proxy support.

Which AI clients are supported?

Any client that supports the Model Context Protocol works: Claude Desktop, Cursor, Windsurf, VS Code, Continue, and more. The same config block works everywhere.
As new MCP-compatible tools ship, Microlink MCP works with them automatically — no updates needed on your end.

What can my AI agent do with it?

Screenshot any URL, generate a PDF, convert a webpage to clean Markdown or HTML, pull normalized metadata and brand logos, scrape custom fields with CSS selectors, extract video and audio sources, collect every link, image, and email, run a Lighthouse audit, detect a site's tech stack, search Google as structured data, and run custom JavaScript in a browser sandbox.
All twenty tools are available through natural language — no code, no API calls, no configuration beyond the initial setup.

Is there a free tier?

Yes. Start immediately with 25 free requests per day — no credit card, no signup required. Every tool except Search is available on the free tier; Search runs on paid plans.
When you need more throughput or pro features, upgrade to a paid plan at any time from microlink.io/#pricing.

How is this different from calling the Microlink API directly?

Calling the API directly means writing HTTP requests, handling auth, parsing responses, and wiring everything up in code. Microlink MCP removes all of that — your AI agent calls the same API through natural language, and structured results come back automatically.
Use the API directly when you need full programmatic control. Use the MCP server when you want your AI assistant to handle web tasks on its own.

How do I install it?

The recommended way requires no installation. Use npx directly in your MCP client config:
mcp.json
{ "mcpServers": { "microlink": { "command": "npx", "args": ["-y", "@microlink/mcp"] } } }
Optionally, install globally to run it as a named command:
Terminal
npm install -g @microlink/mcp
And then reference it via node with the absolute path to the local entry point:
mcp.json
{ "mcpServers": { "microlink": { "command": "node", "args": ["/absolute/path/to/mcp/src/index.js"], "env": { "MICROLINK_API_KEY": "YOUR_MICROLINK_API_KEY" } } } }