MCP Server

One MCP, the whole web.

Name: Spider MCP Server
Author: Spider

Your AI can crawl sites, scrape pages, search the web, and control a real browser. Point your client at mcp.spider.cloud and it works in seconds.

Endpoint Hosted

https://mcp.spider.cloud/mcp

Claude Code Claude Desktop Cursor Windsurf Any MCP client

Get API Key Source on GitHub ↗

01 · Connect

Two ways to set up.

Pick one, paste the config, and you're done.

Claude Code

One command

$ claude mcp add spider \
  --transport http https://mcp.spider.cloud/mcp \
  -H "Authorization: Bearer your-key"

→ added spider · 22 tools

Claude Desktop · Cursor · Windsurf

{
  "mcpServers": {
    "spider": {
      "type": "url",
      "url": "https://mcp.spider.cloud/mcp",
      "headers": {
        "Authorization": "Bearer …"
      }
    }
  }
}

No dependencies · Spider handles crawling, proxies, and browsers

02 · Toolset

22 tools, one connection.

Your AI picks the right one for the job. Here's what it has to work with.

Core 8

spider_crawl
Follow links and pull content from an entire site or section.
spider_scrape
Grab one page. No link following, just the content you asked for.
spider_search
Search the web, then optionally crawl whatever comes back.
spider_links
List every link on a page without downloading anything else.
spider_screenshot
Full-page or viewport screenshot, returned as base64.
spider_unblocker
Get past bot detection walls that would block a normal request.
spider_transform
Turn raw HTML into clean markdown locally. No network call.
spider_get_credits
Check how many credits you have left.

AI 5 Subscription

spider_ai_crawl
Tell it what to crawl in plain English, like "only product pages".
spider_ai_scrape
Describe the data you want and get structured JSON back.
spider_ai_search
Web search that ranks results by what actually matters to your query.
spider_ai_browser
Describe what to do in the browser and it figures out the clicks.
spider_ai_links
Extract links with AI deciding which ones are relevant.

Browser 9

spider_browser_open
Spin up a cloud browser with stealth and anti-bot built in.
spider_browser_navigate
Go to a URL and wait for the page to finish loading.
spider_browser_click
Click something on the page using a CSS selector.
spider_browser_fill
Type into a form field.
spider_browser_screenshot
Take a screenshot of what the browser sees right now.
spider_browser_content
Pull the page HTML or just the visible text.
spider_browser_evaluate
Run arbitrary JavaScript on the page.
spider_browser_wait_for
Pause until a selector appears, a nav finishes, or the network settles.
spider_browser_close
End the session and free up the browser.

03 · Under the hood

Why it actually works.

Markdown, not HTML soup

Nav bars, ads, cookie banners, footers — all stripped. You get clean markdown that uses 60–80% fewer tokens than raw HTML, so the context window goes toward your actual task instead of boilerplate.

Sub-second responses

A single page comes back in under a second. Bulk crawls hit 100K+ pages per second. Every slow tool call is dead time where your agent is waiting instead of thinking.

Anti-bot walls handled

Cloudflare, Akamai, PerimeterX, CAPTCHAs. Spider handles all of it so your agent doesn't burn tool calls retrying blocked requests. 99.9% production success on protected sites.

Real browser, not curl

SPAs, lazy-loaded images, infinite scroll, client-rendered content. Spider runs a real browser so you get the page people actually see, not the empty shell HTTP gives you.

Browser you can drive

Open, navigate, click, type, screenshot, run JS, close. Your agent can log into dashboards, fill forms, or walk through multi-step flows the way a person would.

199+ country proxies

Residential, ISP, mobile, datacenter. Proxies rotate automatically and geo-restricted content just works. Pass a country code if you need a specific location.

04 · In flight

What it looks like at runtime.

Two real examples. You ask a question; the AI picks which tools to call.

Data collection search + crawl

You

Research how Stripe handles idempotency keys and summarize the approach.

Claude

I'll search for Stripe's idempotency documentation and crawl the relevant pages.

spider_search query: "stripe idempotency keys API" · limit: 5

spider_crawl url: stripe.com/docs/api/idempotent_requests · format: markdown

6 pages · 2.3s · $0.006

Stripe uses a header-based idempotency system. Here's how it works:

01 Client sends Idempotency-Key header with a unique UUID.
02 Server caches the response for 24 hours, keyed by that UUID.
03 Retries with the same key return the cached response, not a new charge.

Browser session browser tools

You

Go to news.ycombinator.com, find the top post, and screenshot it for me.

Claude

I'll open a browser, navigate to Hacker News, grab the top story, and take a screenshot.

spider_browser_open

session: abc-123 · browser: chrome

spider_browser_navigate url: news.ycombinator.com

spider_browser_content format: text

spider_browser_click selector: ".titleline a"

spider_browser_screenshot

screenshot captured · 1280×720

spider_browser_close

The top post on Hacker News right now is "Show HN: ...". Here's the screenshot of the linked article:

[screenshot image displayed]

Ship it

Give your AI web access.

One API key. One config change. Your AI can reach the entire web.

Get API Key API docs