logo
FR

Webpage to Markdown

Turn web pages into clean Markdown for AI and RAG: one row per URL with the main content, title, author, publish date and the page's links split into same-site and external; also works on pages that need JavaScript; pages that cannot be read say why and are not billed. Not affiliated with the websites read

Développeur
Tarifs
0,0005 $US / Page read (main content returned)
Dernière mise à jour
2026-11-13

Aperçu

Give it a list of web pages and get back one row per page: the main content as clean Markdown, plus the title, site name, author, publish date and the page's links, already split into same-site and external links.

The first step of almost every AI workflow that touches the web — summarizing an article, answering questions from a docs site, filling a RAG knowledge base — is turning a page into clean text. Raw pages are mostly navigation, ads, footers and cookie banners, which waste context and confuse the model. This app keeps only the main content and converts it to Markdown with headings, lists, tables, links and images intact, so people and models can both read it.

Every URL gets a row with a clear status: read, opened but empty, asked for a human check, refused, disallowed by the site, gone, or not a web page. You never send 10 URLs and get 7 rows back without knowing what happened to the other 3, and pages that could not be read are not charged.

Data notes

  • This app reads public pages at the addresses you give and converts their format; the content belongs to the original websites, so follow their copyright and terms of use. This is not an official product of any of these websites and is not affiliated with them. The fee covers fetching, rendering and conversion, not the content itself.
  • Only the pages you list are read; links on them are not followed, so the number of pages and the cost are known before a run. To go deeper, pick the next pages from "Same-site links" and read them in another run.
  • Pages whose content only appears after JavaScript runs are read too.
  • Sites are treated politely: their robots.txt is respected, and logins, paywalls, CAPTCHAs and interactive checks are never bypassed. Such pages come back with a status instead of content.
  • "Author" and "Published" are only filled when the page marks them explicitly; they are never guessed from the text. Dates are given as year-month-day.
  • Not covered: content behind a login, PDF and other files, and sites that block automated visitors outright (many social networks and large marketplaces). Main content longer than 100,000 characters is cut, and the status details say so.

What the results look like

Each row is one URL.

URLStatusTitlePublishedCharacters
https://www.anthropic.com/news/claude-opus-4-5okIntroducing Claude Opus 4.52025-11-2410342
https://www.allbirds.com/products/mens-tree-runnersokMen's Tree Runner - Jet Black (White Sole)4000
https://stackoverflow.com/questions/231767verification0

Each row also has the Markdown content, site name, author, language, description, same-site links, external links, images, status details and the time it was read.

Use cases

  • Feed an AI: hand "Content (Markdown)" to a model for summaries, Q&A or translation, without menus and ads in the way.
  • Build a knowledge base: store "Content (Markdown)" with "Title", "URL" and "Published" for each page, so answers can cite where they came from.
  • Read a docs site or blog section layer by layer: start with the entry page, choose the next URLs from "Same-site links" by link text or path (for example only addresses under /docs/), and read them in the next run. You decide how deep to go and how many pages each round reads.
  • Clean a URL list: filter on "Status" to find pages that are gone, refuse access or have no content, and fix or drop them; "External links" shows which outside sources each page cites.

Périmètre et limites

Up to 100 URLs per run; only the given pages are read, links on them are not followed. robots.txt is respected; logins, CAPTCHAs and interactive checks are never bypassed. Main content is capped at 100,000 characters per page. Billed per page whose main content was read; pages that could not be read also get a row but are not billed. The fee covers fetching, rendering and conversion, not the content itself

  • Adapté pour
  • You have a list of page URLs and need their main content as clean Markdown for an LLM to read, summarize or answer from, or to load into a RAG knowledge base
  • You need the title, author and publish date of articles or blog posts together with their text
  • You want to read a documentation site or a blog section layer by layer: read the entry page, pick the next URLs from `internal_links` (skipping ones already read), and call again with them; you decide what each round reads, so its cost is known in advance
  • À ne pas utiliser pour
  • You have no URLs and want to find pages by keyword: use a web search app first, then read the URLs it returns here
  • The content is behind a login or a paywall, or the site shows an interactive CAPTCHA: these come back as blocked or verification
  • You need PDF, Word or other files: only web pages are read (they come back as not_html)
  • Social networks and marketplaces that refuse automated access (LinkedIn, Facebook, Reddit, Amazon and similar) usually come back as blocked

Gestion des échecs

Comportement en cas d'échec et de nouvelle tentative déclaré par le développeur. Nous vous recommandons de l'inclure dans votre prompt système lors de l'intégration.

  1. 1Every URL gets exactly one row. `status` says what happened: ok, empty_content, verification, blocked, robots_disallowed, not_found, not_html, timeout, dns_error, error; only ok is billed (`billed_pages` = 1)
  2. 2verification, blocked, timeout and error can be retried later
  3. 3robots_disallowed means the site's robots.txt asks crawlers not to fetch that page; it is not read
  4. 4empty_content means the page opened but has no readable main content (a list, login or app page); its title and links are still returned

Entrée

Paramètres requis pour appeler cette App, générés à partir de l'input.schema du manifest.json.

  • URLsurlsRequisarray<string>
    Web pages to read, 1–100. Each URL gets one row. Only these pages are read; links on them are not followed. A URL given twice is read once.
    Nombrejusqu'à 100 élémentsExemple["https://www.anthropic.com/news/claude-opus-4-5"]

Sortie

Structure des champs d'un enregistrement, générée à partir de l'output.schema du manifest.json.

  • URLinput_urlstring
    The URL as given in the input, to match rows to inputs
    Exemplehttps://www.anthropic.com/news/claude-opus-4-5
  • Final URLfinal_urlstring
    The address actually read; differs from the input when the page redirected
    Exemplehttps://www.anthropic.com/news/claude-opus-4-5
  • Statusstatusstring
    ok = main content read; empty_content = the page opened but has no readable main content; verification = the site asks for a human check; blocked = the site refused access; robots_disallowed = robots.txt does not allow this page; not_found = the page does not exist; not_html = not a web page (e.g. PDF); timeout = no response in time; dns_error = the domain does not exist; error = could not be opened. Only ok is billed
    Exempleok
  • Status detailsstatus_textstring
    Plain-English explanation of the status, with the reason when the page was not read; also notes when the content was cut at 100,000 characters
    ExempleMain content read

Exemple de données

Généré à partir d'output.schema. Les champs sans valeur d'exemple utilisent des espaces réservés structurels.

output.example
{
  "input_url": "https://www.anthropic.com/news/claude-opus-4-5",
  "final_url": "https://www.anthropic.com/news/claude-opus-4-5",
  "status": "ok",
  "status_text": "Main content read",
  "billed_pages": 1,
  "title": "Introducing Claude Opus 4.5",
  "site_name": "Anthropic",
  "author": "Anthropic",
  "published_at": "2025-11-24",
  "language": "en",
  "description": "Introducing Claude Opus 4.5",
  "markdown": "Our newest model, Claude Opus 4.5, is available today...",
  "char_count": 10785,
  "internal_links": "<internal_links>",
  "external_links": "<external_links>",
  "images": "<images>",
  "checked_at": "2026-11-12T10:00:00+00:00"
}

Intégration

Cette App peut être intégrée via MCP, API, SDK ou export de fichiers — tous les canaux partagent les mêmes capacités et la même tarification. Chaque requête s'authentifie avec l'en-tête Authorization: Bearer à l'aide d'une API Key (longue durée, créée dans la console Data Hub) ; les clients MCP peuvent aussi se connecter via OAuth, sans clé. D'autres options comme la CLI et Skill arrivent bientôt. Cette App s'exécute de manière asynchrone et peut être annulée en cours de route : Après l'annulation, l'exécution s'arrête. Les enregistrements déjà livrés sont conservés et facturés.

Avec MCP (Model Context Protocol), vous pouvez appeler cette App directement depuis des clients IA comme Claude et Cursor. Choisissez votre client et votre mode d'authentification, puis copiez la config ci-dessous.

Config du client

Remplacez la valeur après Bearer par votre API Key longue durée. Fonctionne dans n'importe quel client, en CI ou en environnement sans navigateur.

mcpServers config
{
  "mcpServers": {
    "octoparse_datahub": {
      "type": "http",
      "url": "https://mcp-v2.octoparse.com?pin=MagicData/webpage-to-markdown",
      "headers": { "Authorization": "Bearer <YOUR_API_KEY>" }
    }
  }
}

Laissez l'IA tout configurer pour vous

Vous ne voulez pas modifier les configs à la main ? Copiez le prompt d'installation et collez-le dans n'importe quel client IA — il effectuera la configuration à sa manière. (Par défaut, il vous guide pour vous connecter et ne vous demande jamais d'envoyer votre API Key dans la conversation.)

detail.access.mcp.composeHint

Tarifs

Page read (main content returned)

Mesuré en additionnant le champ de sortie billed_pages. Les exécutions échouées ne sont pas facturées.

0,0005 $US/ Page read (main content returned)

Plusieurs événements de facturation se cumulent indépendamment. Voir chaque élément pour les détails. Les exécutions échouées ne sont pas facturées.

Voir les crédits

Essayer maintenant

Renseignez les paramètres et lancez l'exécution — les résultats proviennent d'un appel réel.

Exemple
Les paramètres sont validés par rapport à l'input.schema avant la soumission
0,0005 $US / Page read (main content returned)