Hydrafetch MCP server

Clean web data for developers and agents.

Hydrafetch gives your agent the live web. It reads a page and hands back its main content as clean markdown, with the navigation, banners and boilerplate stripped out, so a page costs your model a few hundred tokens instead of a few thousand. It can also search the web, map and crawl a site, pull fields out of pages as JSON, and read a company's brand, logo and design system from its domain.

This is a hosted, remote MCP server. There is nothing to install or run.


Server URL	`https://api.hydrafetch.com/mcp`
Transport	Streamable HTTP
Sign-in	OAuth in the browser, or your Hydrafetch API key as a Bearer token
Official MCP Registry	`com.hydrafetch/web`

Connect

Most clients only need the URL. Add it, sign in to Hydrafetch in the browser window that opens, choose a workspace, and the tools appear. If you don't have an account yet, you can create one from that same window.

If your client can't run a browser sign-in, or you want to connect a script or CI job, create an API key in the dashboard and send it as Authorization: Bearer <your key>. Each client below shows both ways where it supports them.

Claude (claude.ai and Claude Desktop)

Open Customize, then Connectors.
Click +, then Add custom connector.
Paste https://api.hydrafetch.com/mcp and click Add.
Click Connect and sign in to Hydrafetch.

Custom connectors work on every Claude plan, though the Free plan allows one. On Team and Enterprise plans, an owner adds the connector under Organization settings, Connectors first, and members then connect it from Customize, Connectors.

ChatGPT

Open Settings, then Security and login, and turn on Developer mode.
Go to chatgpt.com/plugins and click the plus button.
Name it Hydrafetch, paste https://api.hydrafetch.com/mcp under Connection, choose OAuth, and create it.
Sign in to Hydrafetch when asked.
In a chat, choose Developer mode from the plus menu and select Hydrafetch.

Developer mode is available on the web for Plus, Pro, Business, Enterprise and Education accounts.

Cursor

Install Hydrafetch from the Cursor Marketplace in Customize, or add it to ~/.cursor/mcp.json to use it in every project (or .cursor/mcp.json to keep it to one repo):

{
  "mcpServers": {
    "hydrafetch": {
      "url": "https://api.hydrafetch.com/mcp"
    }
  }
}

Cursor runs the sign-in the first time it connects. To use an API key instead, add a headers block:

{
  "mcpServers": {
    "hydrafetch": {
      "url": "https://api.hydrafetch.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_HYDRAFETCH_API_KEY" }
    }
  }
}

VS Code

Add this to .vscode/mcp.json in your project, or run MCP: Open User Configuration from the command palette to make it available in every workspace:

{
  "servers": {
    "hydrafetch": {
      "type": "http",
      "url": "https://api.hydrafetch.com/mcp"
    }
  }
}

Start the server from the file or the MCP view, and VS Code asks you to sign in to Hydrafetch.

Cline

Open the MCP Servers panel, go to Remote Servers, and fill in:

Server Name: hydrafetch
Server URL: https://api.hydrafetch.com/mcp
Transport Type: Streamable HTTP

Click Add Server. Cline offers to sign you in when the server asks for it. To use an API key instead, choose Configure and add the entry to cline_mcp_settings.json:

{
  "mcpServers": {
    "hydrafetch": {
      "type": "streamableHttp",
      "url": "https://api.hydrafetch.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_HYDRAFETCH_API_KEY" },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Windsurf

Windsurf is now Devin Desktop. For the Cascade agent, open the ... menu at the top right of the Cascade panel, choose Open MCP config file, and add:

{
  "mcpServers": {
    "hydrafetch": {
      "serverUrl": "https://api.hydrafetch.com/mcp"
    }
  }
}

Save the file and sign in when Cascade connects. To use an API key, add "headers": { "Authorization": "Bearer YOUR_HYDRAFETCH_API_KEY" } next to serverUrl.

For the Devin Local agent, run this in a terminal instead:

devin mcp add -s user hydrafetch https://api.hydrafetch.com/mcp
devin mcp login hydrafetch

Claude Code

claude mcp add --transport http --scope user hydrafetch https://api.hydrafetch.com/mcp

Then run /mcp inside Claude Code and sign in, or run claude mcp login hydrafetch from your terminal. To use an API key instead:

claude mcp add --transport http --scope user hydrafetch https://api.hydrafetch.com/mcp \
  --header "Authorization: Bearer $HYDRAFETCH_API_KEY"

Use --scope project in place of --scope user to share the server with one repo through its .mcp.json.

Any other client

Point it at https://api.hydrafetch.com/mcp over Streamable HTTP. If it supports OAuth with dynamic client registration, it will find everything it needs from that URL. If it doesn't, send Authorization: Bearer <your API key> on every request.

Tools

Tool	What it does	Cost
`scrape`	Reads one page and returns its main content as clean markdown, with optional HTML, links or structured data.	1 credit
`search`	Searches the web and returns ranked results with title, URL and snippet, and can read each result as markdown.	1 credit, plus 1 per result read
`map`	Lists the URLs on a website from its sitemap and links, without fetching their content.	1 credit
`crawl`	Reads the pages of a site as one background job, following links from a starting URL.	1 credit per page
`crawl_status`	Shows a crawl's progress and returns its results a page at a time.	Free
`batch`	Reads a list of URLs you already have as one background job.	1 credit per URL
`batch_status`	Shows a batch's progress and returns its results in the order the URLs were sent.	Free
`extract`	Pulls specific fields out of one or more pages as JSON, shaped by a JSON schema or a prompt.	5 credits per URL
`brand`	Returns a company's logos, colours, name, description and social profiles from its domain.	5 credits
`logo`	Returns a company's logo as an image URL you can embed, for a light or dark background.	1 credit
`styleguide`	Reads a site's design system: colours by role with contrast ratios, type scale, corner radius and button styles.	10 credits
`screenshot`	Renders a page in a real browser and returns a PNG of it.	5 credits
`images`	Collects a page's images with their metadata, without rendering the page.	1 credit

Every tool except crawl and batch only reads, and is marked read-only so clients can run it without asking each time. crawl and batch start a job that spends credits, so most clients will ask before running them. The full input schema for every tool is published at api.hydrafetch.com/mcp/tools, no key needed.

Example prompts

"Read https://example.com/pricing and summarise their plans as a table."
"Find every page on example.com, then read the docs section as markdown."
"Search the web for post-quantum TLS adoption and give me the top five sources with what each one says."
"Extract the name, price and availability from these ten product URLs as JSON."
"What does linear.app do, and what are its brand colours?"
"Take a screenshot of https://example.com on a phone-sized screen."

Pricing

New accounts get free credits on sign-up, with no card needed. Each tool's cost is in the table above, and every result tells your agent how many credits it used and how many are left. A failed call costs nothing, so it is safe to retry. Plans and top-ups are on the pricing page.

A note on what pages say

A page your agent reads was written by someone else, and some pages contain text aimed at the model reading them. Treat what comes back as material to reason about, never as instructions to follow. The untrusted content page in the docs covers this in more detail.

License

The files in this repository are released under the MIT License. The Hydrafetch service itself is covered by its terms.

Hydrafetch MCP server