REST API + MCP server for clean Markdown

Turn any web page, search result, or document into clean Markdown

Scrape a URL, run a web search, or parse a document (PDF, Office documents and more) and get back LLM-ready Markdown with one key. Build with REST. Connect AI agents over MCP.

Start with $0.50 free Explore endpoints

≈ 500 free requests to start
Balance never expires
No subscription

Markdown

https://en.wikipedia.org/wiki/Markdown

sample 0.4s

# Markdown

Markdown is a lightweight markup language for creating
formatted text using a plain-text editor.

## History

Markdown was created in 2004 by John Gruber...

Or parse a document by URL. That fetches the file, so it is metered like a scrape.

Q3 Revenue

quarterly-report.xlsx

sample $0.001

# Q3 Revenue

| Region | Revenue | Growth |
| ------ | ------- | ------ |
| EMEA   | $1.2M   | +14%   |
| APAC   | $0.9M   | +22%   |
| Americas | $2.1M | +8%    |

Uncached searches can take up to 90 seconds. Searching and scraping the results…

sample best seo tools · us

The 20 Best SEO Tools, Reviewed

https://example.com/best-seo-tools

A hands-on comparison of the leading SEO platforms for research, audits, and rank tracking. Each result comes back with the page already scraped to Markdown.

Free SEO Tools Every Marketer Needs

https://example.com/free-seo-tools

Read the content of every result, not just the link, so your agent can reason over the page.

Works with MCP clients and agent workflows you already use

REST API for products

Clean endpoints for apps, dashboards, and backend jobs

Use the API when clean Markdown belongs inside your software. Every endpoint has documentation, a playground, and simple usage-based pricing from $0.001 per call.

scrape

Fetch any URL, web page or document through a real Chrome fingerprint and get back clean Markdown plus the raw HTML.

$0.001 / 1K words Docs & playground

search

Run a web search and get the top results with each result page already scraped to Markdown.

$0.001 + $0.002/result Docs & playground

parse

Upload a document (PDF, Office, EPUB, CSV, XML) and get back clean Markdown. Stateless, up to 25 MB.

$0.001 / 1K words Docs & playground

account_balance

Remaining prepaid balance for the account, in dollars.

Free Docs & playground

POST /api/v1/scrape

example.com

Live web page

Any URL in, clean Markdown out

Real Chrome fingerprint • residential proxy

HTML

raw

markdown

DOC

parsed

JSON Response

{
  "url": "https://example.com",
  "title": "Example Domain",
  "markdown": "# Example Domain\n\nThis domain is for use in\nillustrative examples...",
  "html": "...",
  "metadata": {
    "title": "Example Domain",
    "canonical": "https://example.com/",
    "description": "Illustrative example page"
  }
}

Drag to compare the rendered page and the API response

MCP for AI agents

Add scraping tools to the assistant you already use

MCP (Model Context Protocol) is the open standard that lets AI assistants like ChatGPT and Claude call external tools. One URL connects any MCP client to live scraping and web search, returned as clean Markdown.

MCP server URL

Browser assistants sign in with OAuth; dev tools send your API key. One account, one prepaid balance across REST and MCP.

https://polyscrape.com/mcp

Step-by-step guides for ChatGPT, Claude, Claude Desktop, Claude Code, Cursor, VS Code, Windsurf and Cline.

Full setup guide

Add the free polyscrape Skill — a short playbook that teaches your agent when and how to use these tools well.

Add the Skill

Why polyscrape

Clean Markdown without the plumbing

Start small, build with REST or MCP, and keep usage tied to one prepaid balance.

Pay only when called

No subscription to justify. Every call draws a small usage-based fee from your prepaid balance, so you pay only for the calls you make.

No scraping plumbing

No proxies to manage, no browser farm, no HTML-to-Markdown glue. One key, and clean Markdown comes back.

Scrape, Search, and Parse

Fetch a URL, run a web search, or parse a document in one tool surface.

Live examples

Answers a model can’t give alone.

These are real polyscrape-powered answers assembled from web search and pages scraped straight to Markdown.

Pick a conversation

ChatGPT

live web data via MCP

MCP connected

What are the best running shoes right now, and what do reviewers actually complain about?

Used polyscrape · 2 tool calls

polyscrape.search “best running shoes 2026”

polyscrape.scrape

sources runrepeat.com wirecutter.com reddit.com

Across the top results the same handful of models keep winning, and the complaints are just as consistent as the praise.

1. Three shoes dominate the ranked results, two of them repeat picks across independent review sites, not a single listicle.
2. The most-cited upside is a cushioned daily trainer that handles easy miles and tempo runs alike.
3. The recurring forum complaint is outsole durability past ~400 miles, not comfort.

Pull the key obligations and dates out of this 40-page vendor contract PDF.

Used polyscrape · 1 tool call

polyscrape.scrape “https://example.com/vendor-agreement.pdf”

sources example.com/vendor-agreement.pdf

The PDF comes back as Markdown with headings intact, so the obligations and dates are straightforward to extract.

1. Term is 24 months with auto-renewal unless cancelled 60 days before the end date.
2. Payment is net-30; late fees accrue at 1.5% per month past due.
3. A data-processing addendum in an appendix carries GDPR obligations easy to miss in the raw document.

What is the latest across the top results on this topic today?

Used polyscrape · 1 tool call

polyscrape.search “topic latest news”

sources reuters.com apnews.com techcrunch.com

One enriched search returns the ranked results plus the Markdown of each page, so the agent reads the stories, not just the links.

1. The top results agree on the core facts; the differences are in emphasis and the quotes each outlet chose.
2. Because the page bodies come back inline, the agent can summarize without a second round of fetches.
3. Result depth is caller-controlled, so a quick briefing pulls the top 3 and a deep dive pulls all 10.

Pricing

Top up once. Spend whenever.

New accounts start with a free API balance. Paid balance never expires, so it can sit until your workflows need it.

Free

$0.50 to start

$0.50 free balance to start. Get an API key in seconds with no credit card.

Scrape, Search, and Parse
No credit card required
Perfect for a first integration

Create free account

Starter

$10 one-time

Adds $10 to your API balance for steady production traffic and early automated jobs.

Balance never expires
From $0.001 per call
Useful for small pipelines

Get started

Pro Most popular

$100 one-time

$130 spendable balance, including a $30 bonus. The sweet spot for production workloads.

$100 + $30 free balance
Best value per workload
Built for automated pipelines

Get started

FAQ

Questions before you connect?

Usage-based, charged against your prepaid balance. Scrape and parse cost $0.001 per 1,000 words of Markdown returned (minimum $0.001, capped at $0.10). Search costs $0.001 for the search plus $0.002 per result returned. Prices are identical on REST and MCP.

Yes. Parse is a billable endpoint on the API and MCP, and in the dashboard, priced like scrape: $0.001 per 1,000 words of Markdown returned (minimum $0.001, capped at $0.10). Parsing a document by URL runs through scrape at the same word-based price.

No. Your prepaid balance never expires, so you can use it whenever your workflows need it. In the unlikely event polyscrape is ever discontinued, unused paid balance is refunded.

Paid API requests stop with a 402 until you add more balance. This keeps spend predictable and prevents unexpected charges.

10 requests per second per API key (5/second for parse, which is more CPU-heavy), shared across REST and MCP. Going over returns a 429, which costs nothing; add a short backoff and retry.

Yes. ChatGPT, Claude, Cursor or any MCP client connects to the MCP server; your own code calls the REST API. One key and one prepaid balance cover both.

Because bot-protected pages fight back. We run a real Chrome fingerprint over a residential proxy, convert HTML to Markdown, parse documents, and cache results, all handled. One key gives your agent scrape and search as structured JSON, over MCP or REST.

No. We don't train on your inputs and we don't sell your data. You're free to cache responses.

You get a clean, structured response. A completed request is billable even when the body is empty; failed calls (402, 429) are never charged.

Give your product and agents clean Markdown.

Start with a free key and $0.50 free balance, on REST, MCP, or both.

Start with $0.50 free