Glossary

Model Context Protocol (MCP)

One way for any agent to use any tool — including reading a page that needs JavaScript.

By Roger Campos · Last updated: September 2026

TL;DR

The Model Context Protocol (MCP) is an open protocol, introduced by Anthropic in November 2024, that standardises how AI applications connect to external tools and data. An MCP server offers tools, resources and prompts; an MCP client — Claude, an IDE, an agent framework — discovers and calls them over JSON-RPC.

Start free See the API

Free plan, no credit card. 1,000 credits a month.

How it works

How MCP works

Before MCP, every AI application wired up every tool its own way. MCP puts one protocol between them. A server describes what it offers — tools the model can call, resources it can read, prompts it can use — each with a name, a description and a JSON Schema for its arguments. A client inside the AI application connects, lists them, and lets the model decide when to call one. Messages are JSON-RPC 2.0.

Servers run in one of two ways. A local server is a program the client starts on your machine and talks to over stdin and stdout. A remote server is a URL the client talks to over streamable HTTP, with nothing to install.

In practice

Reading the web through MCP

Fetching a URL is one of the first tools anyone gives an agent. The reference fetch server makes a plain HTTP request and converts the HTML — which is fine for static pages and returns an empty shell for a client-rendered one, because nothing runs the JavaScript. An agent reading the modern web needs a server that renders the page first.

URLpipe

URLpipe's MCP server

URLpipe runs a remote MCP server at https://urlpipe.dev/mcp. Every page is rendered in real Chrome. Its 14 tools are the nine API operations — fetch_markdown, capture_screenshot, lighthouse_audit and the rest — plus get_result, get_request, list_requests, list_projects and get_usage, at the same credits as the HTTP API. It authenticates with an organization token, which can be read-only, limited to one project, or set to expire.

MCP client configuration

{
  "mcpServers": {
    "urlpipe": {
      "type": "http",
      "url": "https://urlpipe.dev/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Tell your agent once that fetch_markdown is the cheap way to read a page (1 credit) and summarize_page is for when the summary is the deliverable (17 credits). Setup per client is in the MCP docs.

Try it

Model Context Protocol (MCP), on a page you choose

Turn any URL into clean Markdown

Paste a link and get the page's main content as tidy Markdown — headings, lists, links and code kept, navigation and cookie banners stripped. The format LLMs and RAG pipelines work best with.

Try it free

FAQ

Frequently asked questions

What is an MCP server?

A program or URL that offers tools, resources and prompts to AI applications over the Model Context Protocol. The application's MCP client lists them and lets the model call them.

What is the difference between a local and a remote MCP server?

A local server is a process the client starts on your machine and talks to over stdio. A remote server is reached over HTTP at a URL, so there is nothing to install.

Can an MCP server read JavaScript-rendered pages?

Only if it renders them. A server that makes a plain HTTP request sees the HTML source; one that loads the page in a headless browser sees what a visitor sees.

Keep reading

Best MCP servers for web browsing Windsurf Why AI agents read empty pages Web access for Claude and Cursor Claude Desktop The URLpipe API as an MCP server All of glossary

See it on your own pages.

Free plan, no card. Confirm your email and your API key is live — you'll be making real requests in minutes.

Start free Email the team