---
title: "Fetch one web page and return its readable text (markdown or plain text) with title and basic metadata."
method: POST
path: "/v2/ops/web.scrape"
tags: ["web"]
---

# Fetch one web page and return its readable text (markdown or plain text) with title and basic metadata.

`POST /v2/ops/web.scrape`

## Request

- Base URL: `https://api.agentkey.app`
- URL: `https://api.agentkey.app/v2/ops/web.scrape`
- Auth: HTTP bearer

Fetches a single URL through a scraping provider and returns the main content as markdown (default) or plain text, plus title, description, language, publish time and the target's HTTP status where the provider reports them. format=text routes to a provider that can deliver plain text.

Fixed price: 0.2 credits per successful call, whichever provider serves it. When firecrawl serves the call the fixed price is the floor: the provider meters the actual cost (per page / per document) and credits_charged reports it; describe carries the provider's note.

## Request body

- object
  - `fields` string[]: Projection: keep only these dotted paths of data, e.g. ["items.id", "items.text"].
  - `format` 'markdown' | 'text': Format of the returned text. markdown keeps headings, lists and links; text is plain text.
  - `include_raw` boolean: Attach the raw upstream payload as `raw` (not a stable contract; truncated above 256 KB).
  - `provider` string: Shorthand for routing.only = [provider].
  - `routing` Routing: Provider routing override. order: try these providers first; only: restrict to these (equivalent to pinning); allow_fallbacks: false tries only the first eligible provider.
    - `allow_fallbacks` boolean: false = try only the first eligible provider.
    - `only` string[]: Restrict to these providers.
    - `order` string[]: Providers to try first, in order.
    - `strategy` 'auto': How eligible providers are ordered. auto = the method's curated order. fastest / stable / popular are specified but not implemented yet (400 until then). The price is fixed whatever the strategy; mutually exclusive with order.
  - `url` string, required: Absolute http(s) URL to fetch.

## Response `200`

Success envelope

- WebPageEnvelope: Success envelope whose data is a WebPage.
  - `attempts` Attempt[], required
    - `error_type` string: validation | not_found | rate_limit | timeout | upstream_unauthorized | upstream_unavailable | provider_error
    - `provider` string, required: Provider slug.
    - `refunded` boolean: true when the reservation for this attempt was refunded.
    - `status` integer, required: HTTP status AgentKey mapped this attempt to (200 on success).
    - `took_ms` integer, required
    - `tool` string, required: Registry tool name (Provider/operation) that was called.
    - `upstream_status` integer: The provider's real HTTP status; absent when nothing was received.
  - `credits_charged` number, required: Credits charged for this call (the method's fixed price).
  - `data` object, required: One fetched web page reduced to readable text, normalized across scrapers.
    - `content_type` string, nullable, required: MIME type of the fetched document (text/html, application/pdf) when the scraper reports it.
    - `description` string, nullable, required: Meta description / summary line when the page has one.
    - `final_url` string, nullable, required: URL after redirects as reported by the scraper; null when the scraper does not report it.
    - `format` 'markdown' | 'text' | 'html', required: Format of `text` as delivered by the scraper.
    - `lang` string, nullable, required: BCP 47 language tag of the page when the scraper reports one.
    - `published_at` string, date-time, nullable, required: Publish time of the page when the scraper extracts one (article meta tags), RFC 3339 UTC.
    - `status_code` integer, nullable, required: HTTP status the target site answered with, when the scraper reports it.
    - `text` string, required: Main content of the page in the requested format (markdown or plain text). Boilerplate (navigation, footers) is removed where the scraper supports it.
    - `title` string, nullable, required: Page title (document title or og:title).
    - `url` string, required: The URL that was requested (echoed from the input).
  - `op` string, required: Method name.
  - `op_version` integer, required: Method version that served the call.
  - `provider` string, required: Provider that actually served the call.
  - `raw` unknown
  - `raw_truncated` boolean
  - `request_id` string, required: Correlation id; quote it in support requests.
  - `took_ms` integer, required
  - `unsupported_fields` string[], required: Output paths this provider structurally does not supply; such fields are null for a different reason than 'missing on this record'.

## Other responses

- `400`: Validation error (unknown field, bad cursor, bad routing)
- `402`: Insufficient credits
- `404`: Not found (method, or upstream object)
- `429`: Rate limited upstream
- `502`: All providers failed
- `504`: Timed out

---

[API](https://skmtc.dev/agentkey/apis/agentkey-scenario-methods.md) - [All operations](https://skmtc.dev/agentkey/apis/agentkey-scenario-methods/llms.txt) - [OpenAPI document](https://skmtc.dev/agentkey/apis/agentkey-scenario-methods/revisions/a7cf5b682fdd?raw)