hydrafetch
Get started

[ Integrations ]

One MCP endpoint. Every client.

13 tools, one URL, nothing to run locally.

The endpoint

https://api.hydrafetch.com/mcp

A streamable HTTP MCP server. Authenticate with an API key as a bearer token, or let the client run the OAuth flow and sign in through the browser.

Get an API key

scrape

Turn a URL into clean, LLM-ready markdown and structured data.

1 credit

map

Enumerate a site's URLs from its sitemap and links, without scraping them.

1 credit

crawl

Discover and scrape a whole site as one job. Returns a crawl id straight away; read it with crawl_status. Use this instead of calling scrape in a loop.

1 credit per page scraped

crawl_status

Read a crawl started with crawl: how far it has got, and the pages it has scraped.

free

batch

Scrape a list of URLs you already have as one job. Returns a batch id straight away; read it with batch_status.

1 credit per URL scraped

batch_status

Read a batch started with batch: how far it has got, and the results so far.

free

search

Search the web and return ranked results: title, url and snippet. Set scrapeResults to also fetch each result as markdown, at 1 extra credit per page.

1 credit plus 1 per scraped result

brand

A company's brand from its domain: logos for light and dark backgrounds, its real colours, name, description and socials. Answers 'what is this company' far better than a raw page dump.

5 credits

logo

A company's logo as a directly embeddable image URL, chosen for the background you name. Use this rather than brand when the mark is all you need — it costs a fifth as much and returns one asset instead of the whole record.

1 credit

extract

Pull typed structured data from one or more URLs by JSON schema or prompt.

5 credits per URL

styleguide

A site's design system read from computed styles in a real browser: colours by role with contrast ratios, the type scale, corner radius and button styling. Values defined through CSS variables come back resolved to the hex actually painted.

10 credits

screenshot

Render a page in a real browser and capture a PNG, returning its public URL. Use when the question is what the page looks like rather than what it says.

5 credits

images

Harvest a page's images with their metadata, without rendering it. Cheaper than a screenshot and returns the source images rather than a picture of the page.

1 credit

[01 / Clients]

Paste one block and restart.

Sign in through the browser, or send your key as a bearer token.

Claude Code

Scoped to your user, so it is available in every project. Swap --scope user for --scope project to commit it with one repo instead.

With an API key
Terminal
claude mcp add --scope user --transport http hydrafetch https://api.hydrafetch.com/mcp \
  --header "Authorization: Bearer $HYDRAFETCH_API_KEY"
Or sign in instead
Terminal
claude mcp add --scope user --transport http hydrafetch https://api.hydrafetch.com/mcp

# then run /mcp inside Claude Code and sign in

[02 / Official clients]

One install, every endpoint.

A thin wrapper over the same HTTP surface. Reads HYDRAFETCH_API_KEY, retries what is worth retrying, polls crawl and batch jobs.

Node

$npm install @hydrafetch/node-sdk
npm Source

Python

$pip install hydrafetch
PyPI Source

Go

$go get github.com/Hydrafetch/go-sdk
pkg.go.dev Source

Ruby

$gem install hydrafetch
RubyGems Source

Rust

$cargo add hydrafetch
crates.io Source

PHP

$composer require hydrafetch/hydrafetch
Packagist Source

Two more for the browser, built for a publishable key rather than a secret one. Logo pulls are metered on their own allowance and never touch your credit balance.

[03 / Frameworks]

Or call it from your own pipeline.

HTTP, not MCP. LangChain and LlamaIndex ship their own package; everything else calls an official client.

LangChain

A document loader that hands your splitter clean Markdown instead of the tag soup a raw HTTP loader returns. One page, a whole crawl, or a site map, and a retriever that searches the web.

$pip install langchain-hydrafetch
PyPI Source
loader.py
from langchain_hydrafetch import HydrafetchLoader, HydrafetchSearchRetriever

# One page, a whole site, or just its URLs.
docs = HydrafetchLoader("https://example.com/article").load()
docs = HydrafetchLoader("https://example.com", mode="crawl", params={"limit": 50}).load()

# Stream a large crawl instead of holding it all in memory.
for doc in HydrafetchLoader("https://example.com", mode="crawl").lazy_load():
    index.add(doc)

# Search the web as a retriever.
retriever = HydrafetchSearchRetriever(k=5)
docs = retriever.invoke("best open source vector databases")

LlamaIndex

A reader, so pages arrive stripped of navigation and boilerplate before they ever reach your index. Map a site first and load what you want: crawling everything and discarding most of it is the usual way to waste credits.

$pip install llama-index-readers-hydrafetch
PyPI Source
reader.py
from llama_index.core import VectorStoreIndex
from llama_index.readers.hydrafetch import HydrafetchReader

reader = HydrafetchReader(params={"onlyMainContent": True})

# A single URL, a list of them, or a whole site.
docs = reader.load_data("https://example.com/article")
docs = reader.load_data(["https://example.com/a", "https://example.com/b"])
docs = reader.load_data("https://example.com", mode="crawl", limit=50)

index = VectorStoreIndex.from_documents(docs)

Vercel AI SDK

A tool your model can call mid-conversation when it needs a page it has not seen.

$npm install @hydrafetch/node-sdk
npm Source
tools.ts
import { Hydrafetch } from '@hydrafetch/node-sdk';
import { tool } from 'ai';
import { z } from 'zod';

const hf = new Hydrafetch(process.env.HYDRAFETCH_API_KEY);

export const readPage = tool({
  description: 'Read a web page and return it as clean Markdown.',
  parameters: z.object({ url: z.string().url() }),
  execute: async ({ url }) => hf.markdown(url),
});

[04 / Second opinions]

Do not take our word for it. Ask an AI to check.

Point your coding agent at your own repository, or ask an assistant to compare us against the alternatives. Both answer with things we did not write.

Ask your coding agent

It reads your repo, not our pitch.

Paste one line into any agent that reads your repository. It comes back with file paths, a credit estimate, and where we are not worth it.

Claude CodeCursorCodexAmpAny MCP client
Read the skill first

The skill reads our live endpoint list when it runs, so it covers anything we ship after you paste this.

Ask an assistant

Let it compare us to the alternatives.

The question is already written, and it asks for the comparison rather than the sales pitch. It opens in a new tab with the prompt filled in.

Ask ChatGPTAsk ClaudeAsk GeminiAsk PerplexityTalk to a human

Assistants read the live site, so they answer from what is published today rather than from anything we hand them.

[05 / Questions]

Before you wire it up.

The longer answers live in the docs.

Either. Most clients speak the MCP OAuth flow, so you point them at the endpoint and sign in through the browser. The rest send an API key as a bearer token. The two are equivalent in what they can do.

The same as the matching HTTP endpoint. A page is one credit whatever it took to fetch, extract and brand are five, and a failed call costs nothing. Connecting over MCP never costs more than calling the API directly.

An API key belongs to the workspace, so a key in a shared config works for everyone and the usage lands on the same balance. Signing in with OAuth ties the connection to you inside that workspace.

Almost certainly. The endpoint is a standard streamable HTTP MCP server, so any client that can take a URL and an Authorization header can use it. The listed configs are just the shapes each client expects.

Every connection you authorize shows up on the MCP page in your dashboard, with the client that opened it and when it was last used. Revoke it there and that client stops working immediately. Key-based connections stop when you roll the key.

[ Start ]

Point your agent at it in under a minute.

13 tools, one endpoint, and the same credit cost as calling the API directly.

Get API key Read the docs

Success rate

 

Median scrape

 ms