Turn website content into structured JSON

Give your Agent a starting website URL and the facts you need. Extract them into a custom JSON structure, with controls for page scope, cache reuse and whether values must be stated on the page.

Your AI

Use with your AI

For tasks with several steps, continue in the AI you already use.

CodexClaude CodeCursorWindsurfCline

View instructions

Use extract-structured-data-context-dev to read the page the user shares and return the details they ask for. Accept the page link and a plain-language description of the requested details; never ask them to write JSON, code, or a technical field list. Prepare the result format yourself. When the user asks for facts stated on the page, use only supported details and leave unknown items blank. Read only the shared page unless the user asks to follow its links; if that is unclear, ask a simple question. Reuse the existing Toolmu connection and authorization, follow the on-page setup guide if needed, and stay within the agreed budget and current price. Include source page links and distinguish missing information from a confirmed empty result. Do not claim complete website coverage or fresh content without evidence.

Task:
From [company introduction page link], find the company name and a link to its contact page. Use only information stated on that page; leave details blank if they are not listed.

Paste into your connected AI's conversation.

Connection guide

Up to 0.0108 credits / call

The final charge is based on confirmed usage. Unused reserved credits return to your available balance.

Context.dev

Turn website content into structured JSON

Give your Agent a starting website URL and the facts you need. Extract them into a custom JSON structure, with controls for page scope, cache reuse and whether values must be stated on the page.

You receive

Receive an extracted data object shaped by your JSON Schema, accompanied by analyzed page URLs, crawl counts and cache context. Your Agent can organize the requested facts for comparison or further processing and point out missing values.

Start with your task

From [company introduction page link], find the company name and a link to its contact page. Use only information stated on that page; leave details blank if they are not listed.

What to know

Page, depth and time limits bound the reading scope without guaranteeing complete site coverage. factCheck true requires page-text grounding; the default allows reasonable inference. Neither independently verifies real-world truth. Set an acceptable cache age and review null or empty values.

Inputs & outputs

Inspect the current inputs, availability, and maximum charge before starting.

Input schema

Input schema
Field / Path	Type	Requirement	Meaning
Extraction settings`body`	`object`	Required	Required settings for extraction: a starting URL and a JSON Schema, plus optional guidance and page, timing and cache limits. Required fields `url, schema`
Starting URL`body.url`	`string`	Required	Required website URL to start reading and extracting from. Format `uri`
Requested JSON Schema`body.schema`	`object`	Required	Required JSON Schema describing the object to return. Specify the properties, types and any nullable values you need; your Agent can draft it from your requirements.
Maximum cache age`body.maxAgeMs`	`integer`	Optional	Explicit maximum age for reused results, in milliseconds, from 0 to 2,592,000,000 (30 days). Set your acceptable age; 0 requests a fresh fetch. Minimum `0` Maximum `2592000000`
Maximum link depth`body.maxDepth`	`integer`	Optional	Maximum link depth from the starting URL. Set 0 to read only the starting page; when omitted, depth is unlimited within the other page and time limits. Minimum `0` Maximum `9007199254740991`
Maximum pages`body.maxPages`	`integer`	Optional	Maximum of 1–50 pages to analyze, default 5. This is an upper limit, not a guaranteed page count or full-site coverage. The current contract bills per call regardless of this limit. Minimum `1` Maximum `50`
Require page grounding`body.factCheck`	`boolean`	Optional	When true, values must be grounded in text stated on the page, with unsupported fields null or empty. False, the default, permits reasonable inference. This checks page grounding, not independent real-world truth.
Supplier timeout`body.timeoutMS`	`integer`	Optional	Supplier deadline in milliseconds, from 1,000 to 300,000. Set it below the Tool request timeout. Context.dev describes HTTP 408 when this deadline is exceeded. Minimum `1000` Maximum `300000`
Wait after page load`body.waitForMs`	`integer`	Optional	Extra browser wait before capturing content, from 0 to 30,000 milliseconds, for pages whose content needs time to render. Minimum `0` Maximum `30000`
Crawl time budget`body.stopAfterMs`	`integer`	Optional	Soft time budget for the crawl phase, from 10,000 to 110,000 milliseconds, default 80,000. It bounds the crawl effort without guaranteeing every requested page is read. Minimum `10000` Maximum `110000`
Extraction guidance`body.instructions`	`string`	Optional	Optional guidance, up to 2,000 characters, on which facts to prioritize and how to interpret ambiguous fields. Maximum length `2000`
Include iframe contents`body.includeFrames`	`boolean`	Optional	Include iframe content in the Markdown given to the extractor. Default false.
Follow subdomains`body.followSubdomains`	`boolean`	Optional	Follow links on subdomains of the starting URL’s domain. Default false; page and time limits still apply.
Wait for animations`body.settleAnimations`	`boolean`	Optional	Wait briefly for animations to settle before reading each page. Default false.

View raw JSON

{
  "type": "object",
  "required": [
    "body"
  ],
  "properties": {
    "body": {
      "type": "object",
      "$schema": "https://json-schema.org/draft/2020-12/schema",
      "required": [
        "url",
        "schema"
      ],
      "~standard": {
        "vendor": "zod",
        "version": 1,
        "jsonSchema": {}
      },
      "properties": {
        "url": {
          "type": "string",
          "format": "uri",
          "description": "Starting website URL to crawl and extract from."
        },
        "schema": {
          "type": "object",
          "description": "JSON Schema describing the object to return. Generate it from a Zod or Pydantic model, or hand-write it — Context.dev fills exactly this shape.",
          "propertyNames": {
            "type": "string"
          },
          "additionalProperties": {}
        },
        "maxAgeMs": {
          "type": "integer",
          "maximum": 2592000000,
          "minimum": 0,
          "description": "Reuse a cached result younger than this many milliseconds. Default 86400000 (1 day), max 2592000000 (30 days). Set 0 to always fetch fresh."
        },
        "maxDepth": {
          "type": "integer",
          "maximum": 9007199254740991,
          "minimum": 0,
          "description": "Maximum link depth from the starting URL (0 = only the starting page). Unlimited when omitted."
        },
        "maxPages": {
          "type": "integer",
          "maximum": 50,
          "minimum": 1,
          "description": "Maximum number of pages to analyze. Default 5, hard cap 50. Does NOT change the price — extraction is billed per call."
        },
        "factCheck": {
          "type": "boolean",
          "description": "When true, every returned value must be grounded in text stated on the page and unsupported fields come back null/empty. When false (default), reasonable inferences are allowed while verifiable specifics stay faithful to the source."
        },
        "timeoutMS": {
          "type": "integer",
          "maximum": 300000,
          "minimum": 1000,
          "description": "Upstream timeout in milliseconds (max 300000). Context.dev aborts the call with a 408 when exceeded. Keep it below the endpoint's requestTimeoutMs so the provider answers before the platform's own budget expires."
        },
        "waitForMs": {
          "type": "integer",
          "maximum": 30000,
          "minimum": 0,
          "description": "Extra browser wait in milliseconds after page load before the content is captured (0-30000). Useful for JavaScript-heavy pages."
        },
        "stopAfterMs": {
          "type": "integer",
          "maximum": 110000,
          "minimum": 10000,
          "description": "Soft time budget for the crawl phase in milliseconds (10000-110000, default 80000)."
        },
        "instructions": {
          "type": "string",
          "maxLength": 2000,
          "description": "Extraction guidance: which facts to prioritize, how to interpret ambiguous fields."
        },
        "includeFrames": {
          "type": "boolean",
          "description": "Include iframe contents in the Markdown handed to the extractor. Default false."
        },
        "followSubdomains": {
          "type": "boolean",
          "description": "Follow links on subdomains of the starting URL's domain. Default false."
        },
        "settleAnimations": {
          "type": "boolean",
          "description": "Wait briefly for animations to settle before each page is read. Default false."
        }
      },
      "additionalProperties": false
    }
  },
  "additionalProperties": false
}

Result schema

No fixed output contract is available. Field types reflect saved result evidence; individual responses can differ.

Result schema
Field / Path	Type	Meaning
Extracted data`result.data`	`object`	The extracted object whose properties depend on your supplied JSON Schema. Read its values with your factCheck setting and any missing or null fields.
Requested title value`result.data.title`	`string`	A title value when the requested JSON Schema includes a title property. This is an example of a requested property, not a fixed field of every extraction.
Starting URL`result.url`	`string`	Starting URL associated with the extraction result.
Extraction status`result.status`	`string`	Extraction status reported by Context.dev, such as ok. Read the extracted fields and crawl statistics as well.
Analyzed URLs`result.urls_analyzed`	`array`	The list of page URLs reported as analyzed for this extraction.
Crawl statistics`result.metadata`	`object`	Reported processing counts and maximum crawl depth for the pages considered.
Reported URL count`result.metadata.numUrls`	`number`	URL count in the crawl statistics; this is not the website’s total page count.
Successful URLs`result.metadata.numSucceeded`	`number`	Count of URLs reported as successfully processed.
Failed URLs`result.metadata.numFailed`	`number`	Count of URLs reported as failed during processing.
Blocked URLs`result.metadata.numBlocked`	`number`	Count of URLs reported as blocked during processing.
Skipped URLs`result.metadata.numSkipped`	`number`	Count of URLs reported as skipped during processing.
Maximum reached depth`result.metadata.maxCrawlDepth`	`number`	Maximum link depth reported for the crawl, relative to the starting URL.
Cache details`result.cache_metadata`	`object`	Cache reuse details associated with the extraction.
Cache age`result.cache_metadata.age_ms`	`number`	Age of the reused result in milliseconds. It is a duration, not the source page’s update time.
Cache status`result.cache_metadata.status`	`string`	Cache status reported for this result, such as hit.

Questions about this tool.

Inputs, results and details to know before running it.

Get connected

Yes. Explain the fields you need and let your Agent draft an object schema, then review the property types and any required or nullable fields. The request must include both the starting URL and the JSON Schema. Output properties are determined by that schema.

Turn website content into structured JSON

Turn website content into structured JSON

You receive

Start with your task

What to know

Inputs & outputs

Input schema

Result schema

Use Toolmu with your AI assistant

Say what you want to find

Check the information found

Questions about this tool.

You receive

Start with your task

What to know

Inputs & outputs

Input schema

Result schema

Use Toolmu with your AI assistant

Say what you want to find

Check the information found

Questions about this tool.

Can my Agent prepare the JSON Schema?

How do I restrict extraction to one page?

What does the page limit control?

Does factCheck verify that the website is correct?

How can I tell which pages contributed?