Turn a web page into markdown for your LLM

Turn any web page into clean markdown for an LLM or RAG pipeline with one API call or MCP tool. Firecrawl and Jina Reader side by side, priced per page.

Ask your agent

Read these five competitor pricing pages as markdown and give me a table of their plans and prices.

Paste this into Claude, ChatGPT or any agent connected to Superpowers.

Failed calls are freeReady to use from 2 sourcesWorks with your AI agent

What you get backFrom a real call on 10/09/2026 · Firecrawl

Firecrawl

Turn the Wikipedia article on the Model Context Protocol into markdown.

Title: Model Context Protocol - Wikipedia.
76,696 characters of markdown in 551 ms, starting at the article itself ("From Wikipedia, the free encyclopedia"), with tables and links kept.
Charged $0.00494 for the page.

Markdown cut to the first 420 characters.

Jina Reader

Same page, with Jina Reader.

87,998 characters in 2,174 ms for $0.00065, about 8 times cheaper.
The extra characters are Wikipedia's menus ("Main menu", "move to sidebar"), which Firecrawl left out.

From a real call on 10/09/2026 · the same comparison run

What comes back

The fields every call returns, in the same shape for every provider.

titleThe page title

markdownThe page content as markdown: headings, lists, tables and links

charsLength of the markdown, to budget tokens before you send it to a model

Providers

Same input and output for every provider. Success rate over 30 days shows after 5 real calls.

Provider	Price per call	Success	Status
Firecrawl	$0.00494	86%	Available
Jina Reader	$0.00065	99%	Available

Frequently asked questions

Firecrawl or Jina Reader?

In our comparison on the same Wikipedia article, Firecrawl answered in 551 ms with the article only (76,696 characters) for $0.00494, and Jina Reader took 2,174 ms, kept the site menus (87,998 characters) and cost $0.00065. Pick Firecrawl when clean text matters, Jina Reader when price matters most.

Can I compare them on my own page?

Yes. POST /v1/compare with category scrape.markdown runs both on the same URL and returns each result, its speed and its price, so you can choose with your own pages.

Does it work on pages that need JavaScript or a login?

Pages that render with JavaScript usually work. Pages behind a login, a paywall or a bot block will fail, and failed calls are not charged.

How do I give this to Claude, Cursor or ChatGPT?

Add the Superpowers MCP server once. Your agent then sees the readers as tools and can turn any URL into markdown by itself.

What does it cost?

Per page, at the prices in the table below. New accounts get $5 of credit, enough for about a thousand pages with Firecrawl.

For developersAPI request, full response, returned fields and MCP setup

Pass a URL and get back the page title and its content as markdown, ready to put in a prompt, a RAG index or an agent's memory. Two readers sit behind the same input and output: Firecrawl and Jina Reader. You pay per page, only when the call works, and your agent can call either one through the Superpowers MCP server or POST /v1/run.

Request

curl -X POST https://superpowers.tools/v1/run \
  -H 'Authorization: Bearer $SUPERPOWERS_KEY' \
  -H 'content-type: application/json' \
  -d '{"tool":"firecrawl.scrape","idempotency_key":"example-1","input":{"url":"https://en.wikipedia.org/wiki/Model_Context_Protocol"}}'

Normalized response

Call made on 10/09/2026 through Firecrawl, answered in 551 ms. Markdown cut to the first 420 characters.

{
  "title": "Model Context Protocol - Wikipedia",
  "markdown": "[Jump to content](https://en.wikipedia.org/wiki/Model_Context_Protocol#bodyContent)\n\nFrom Wikipedia, the free encyclopedia\n\nProtocol for communicating between LLMs and applications\n\n| [![](https://thumb.wikimedia.org/wikipedia/commons/thumb/f/fe/Model_Context_Protocol_logo.svg/250px-Model_Context_Protocol_logo.svg.png?utm_source=en.wikipedia.org&utm_campaign=parser&utm_content=thumbnail)](https://en.wikipedia.org/wik …",
  "chars": 76696
}

MCP configuration

Add the MCP server to Claude, Cursor or another client and ask in plain language.

Claude Code

claude mcp add --transport http superpowers https://superpowers.tools/mcp \
  --header "Authorization: Bearer $SUPERPOWERS_KEY"

Cursor (.cursor/mcp.json)

{
  "mcpServers": {
    "superpowers": {
      "url": "https://superpowers.tools/mcp",
      "headers": {
        "Authorization": "Bearer ${env:SUPERPOWERS_KEY}"
      }
    }
  }
}

Codex

codex mcp add superpowers --url https://superpowers.tools/mcp \
  --bearer-token-env-var SUPERPOWERS_KEY

Start with $5 of credit

Enough for 1,012 calls. No card needed to start.

Create an account