MCP TOOLcubeshell.web.scrape

Web Scrape

Extracts bounded readable Markdown and basic metadata from one exact public page. It does not search the web, crawl a site, sign in, or access private network addresses.

Price

A successful valid page costs 1.50 Cubeshell credits. A definitive failure is not charged. Cubeshell returns the price and exact receipt with every call.

Request

Parameter	Type	Description
`request_id`required	`string`	Stable idempotency key. Replay the same input with the same ID to avoid a second call or charge.
`quote_only`required	`false`	Must be false in the current direct-execution contract.
`input.url`required	`public HTTP(S) URL`	One page, up to 2,048 characters. Credentials, fragments, private hosts and nonstandard ports are rejected.
`max_microcredits`	`integer`	Optional ceiling. Set 1500000 to permit this tool's current 1.50-credit price and reject any higher price before execution.

tools/call

{
  "name": "cubeshell.web.scrape",
  "arguments": {
    "request_id": "scrape-product-page-001",
    "quote_only": false,
    "max_microcredits": 1500000,
    "input": { "url": "https://example.com/" }
  }
}

Result

Parameter	Type	Description
`result.url`	`string`	Normalized requested URL.
`result.final_url`	`string`	Final public URL reported after normal redirects.
`result.content`	`Markdown`	Readable content, bounded to 200,000 bytes.
`result.title`	`string`	Page title when available.
`result.description`	`string`	Page description when available.
`result.http_status`	`200–299`	Canonical page response status.
`result.truncated`	`boolean`	Whether the canonical content was deliberately shortened.

Public pages only

Do not send secrets, cookies, authorization headers, intranet URLs, or logged-in pages. The tool has no input fields for them and fails closed on non-public targets.

Job Listings

Web Screenshot