Crunchbase Scraper

Scrape Crunchbase News articles and the organization links inside them. Export Crunchbase news data to JSON or CSV for funding coverage, market monitoring, and organization mention tracking. This actor uses the public news feed interface; it does not retrieve private organization profile fields.

What data can you extract from Crunchbase?

Each result is one article. The article's organization links are collected into aligned arrays, so a downstream process can join a story to the profile URLs it cites. These are mentions in editorial coverage, not company profile records. If an article names a business without linking to its profile, that name will not appear in the organization arrays.

Field	Description	Example
articleId	Stable numeric identifier for deduplication	94105
articleUrl	Canonical link to the story	A news article URL
title	Headline converted to plain text	“Jumbo-Sized Series A Rounds Are On The Rise”
summary	Editorial excerpt converted to plain text	A short article synopsis
publishedAt	UTC publication timestamp	2026-09-23T11:00:38Z
modifiedAt	UTC last-update timestamp	2026-09-22T15:20:26Z
author	Public byline, when supplied	A byline string
categorySlugs	Editorial categories	["startups", "venture"]
tagSlugs	Editorial tags	[]
imageUrl	Featured image link, when available	An image URL
readingTimeMinutes	Estimated reading time at 220 words a minute	3
companyPermalinks	Distinct linked organization profile URLs, in article order	An array of profile links
companyNames	Visible link labels aligned to the profile URLs	An array of link labels
companyCount	Number of distinct linked profiles	4
text	Full plain article text when requested	Article body

The profile URLs are normalized to one host and lowercase slugs. This makes them useful as join keys across articles. Names come from the article's visible link labels and can vary between stories. Use the permalink for matching; use the label for display. companyCount counts distinct organization profile links, not every business discussed in the article.

The article row remains compact by default. Full text is optional because it can be much longer than the summary. The summary usually supplies enough context for a feed, alert, or review queue. When you need to extract your own signals from the article body, enable includeText.

How to scrape Crunchbase

1. Open the actor and leave the inputs empty to collect the newest public news articles. The default limit is 100 results.
2. Add a phrase in search, a category slug, or a publication date window to narrow the list. You can also enter organization URLs or slugs to find articles linking to those profiles.
3. Set maxItems to the number of article rows you want. A small value is useful while checking a filter.
4. Run the actor. Each result is written to the default dataset as soon as it is processed.
5. Download the dataset as JSON or CSV, or read it through the platform's dataset interface.

For a company mention search, enter the organization's public profile URL if you have it. The actor searches news content using the profile slug, then checks each returned article for an actual link to that profile. This second check prevents a simple text match from being mistaken for a linked mention. Searches are run separately for multiple URLs and article IDs are deduplicated across them.

Input

Parameter	Type	Default	Description
companyUrls	String list	Empty	Organization profile URLs or slugs to find in linked article references.
search	String	Empty	Word or phrase to find in news articles. With organization inputs, the phrase is checked against matched articles.
category	String	Empty	Exact editorial category slug, such as venture or startups.
dateFrom	Date string	Empty	Earliest publication date, inclusive, in YYYY-MM-DD format.
dateTo	Date string	Empty	Latest publication date, inclusive, in YYYY-MM-DD format.
includeText	Boolean	false	Add the full article body as plain text.
maxItems	Integer	100	Stop after this many article rows; allowed range is 1 to 4,000.
proxyConfiguration	Object	Platform proxy	Connection settings for the public news interface.

To collect a bounded news window, use an input like:

{
  "category": "venture",
  "dateFrom": "2026-09-01",
  "dateTo": "2026-09-30",
  "maxItems": 40
}

For a recurring feed, keep the category and search phrase fixed, then adjust the date window for each run. The actor itself does not keep a history of earlier runs. Deduplicate successive datasets by articleId in your own workflow. For an organization-focused workflow, set companyUrls to one or more public profile links or slugs and use articleUrl as the article reference in an alert.

Inputs select news coverage. A profile URL in companyUrls is a filter, not an instruction to scrape that profile. This distinction matters for anyone expecting funding totals, employee counts, industry tags, investment history, or contact details from a profile page. Those fields are not returned by this actor.

Output example

The following selected fields come from one item in a successful 40-article run. Long fields and the arrays of individual profile links are omitted from this display; the dataset contains the full item.

{
  "articleId": 94105,
  "title": "Jumbo-Sized Series A Rounds Are On The Rise",
  "publishedAt": "2026-09-23T11:00:38Z",
  "modifiedAt": "2026-09-22T15:20:26Z",
  "tagSlugs": [],
  "readingTimeMinutes": 3,
  "companyCount": 4
}

In that run, all 40 requested articles had an identifier, URL, headline, summary, publication timestamp, and at least one linked organization profile. This is a measured result for that run, not a guarantee for every date range or category. The set of editorial links depends on what each article contains.

Use cases

• Funding coverage review: Collect articles from the venture category, sort them by publication time, and review the linked organization profiles alongside each headline.
• Market monitoring: Search for a topic and keep a dated archive of article summaries, categories, and linked profiles. This works well when a team wants a repeatable view of a sector rather than a manual reading list.
• Organization mention alerts: Supply a profile URL or slug to find articles that link to it. Store articleId between runs so only unseen coverage triggers a notification.
• Research handoff: Share a CSV of stories with article URLs and organization counts, then open the source article for context before making decisions.
• Content indexing: Use the stable article ID, category slugs, timestamps, and optional text to build a searchable internal index of public news coverage.
• Relationship mapping: Split companyPermalinks into a second table keyed by articleId. This yields a simple article-to-organization relationship table without guessing organization names from prose.

These workflows use editorial references as evidence that an article linked to a profile. They do not establish that a company raised a stated amount, participated in a transaction, or currently has a given operating status. The story itself provides the context for any such conclusion.

How much does it cost to scrape Crunchbase?

The configured result price is $6.40 per 1,000 article rows, or $0.0064 per row. A result is charged when it is delivered to the dataset. The examples below count result charges only; check the run page for the final charge and any account-specific usage details.

Delivered articles	Result charge
40	$0.256
100	$0.64
1,000	$6.40

The golden run delivered 40 rows and finished successfully. Query breadth affects how many pages the actor requests before reaching your limit. A narrow phrase or organization filter may scan returned posts that do not pass the final link check, so the same number of delivered rows can take different amounts of time. Set maxItems to the rows you actually need.

The price is per article, not per linked organization. An article with several profile links is still one result row. If you need one line per organization mention in a spreadsheet, expand the arrays after export; that transformation does not change how the actor bills its dataset rows.

Tips for best results

Use a category slug when the topic aligns with an editorial section. Slugs are exact; a display label may not work. A broad search without a date filter returns the latest matches first. For a historical collection, split the timeline into monthly or quarterly windows and combine datasets by articleId. This helps keep each run bounded and makes retries straightforward.

Use a profile permalink when monitoring an organization. A slug is accepted as a shortcut, but the permalink makes the intended target easier to review. The actor searches the news interface using words from the slug, then accepts an article only if its body links to that profile URL. Editorial coverage can refer to an organization without linking its profile, and such an article will not pass the final filter.

Leave includeText off for lightweight feeds. Turn it on when you need full prose for keyword checks or manual review. The plain-text conversion removes markup and leaves readable spaces between paragraphs, but it does not rewrite editorial content. The summary remains available either way.

For scheduled runs, choose an overlapping date window rather than relying on the exact clock time of the previous execution. Articles can be edited after publication, while their original publication date remains the basis of the date filter. Deduplicate by ID and use modifiedAt if you want to detect updates to an existing article.

When exporting to CSV, arrays may appear as serialized values in cells. If you need one row per organization link, expand companyPermalinks and the corresponding companyNames values in a downstream step. The two arrays have the same order and length.

Limits

This actor does not return organization profile records, funding-round records, investment amounts, investor lists, employee ranges, or profile predictions. The main organization and Discover pages, related internal data routes, and the public profile interface did not provide anonymous access during development. The news endpoint is a separate public source and supplies article content plus links to profiles, not the profile data behind those links.

The news interface returns at most 40 posts per requested page in this actor. The actor checks up to 100 pages per search and respects maxItems. An organization search can return fewer rows than requested because only articles with an exact profile link pass the final filter. Search results are controlled by the news site's index, so a story that is not indexed for the slug may be missed.

Category and tag labels are resolved from the public taxonomy lists. If a list is temporarily unavailable or a term falls outside the first 100 returned entries, that slug can be absent from the row even though the article still appears. Publication and modified times come from the source in UTC. Some older articles may omit a featured image, byline, tags, or linked organization profiles.

The readingTimeMinutes field is an estimate based on the extracted word count. It is meant for display and scheduling a reading queue, not as an editorial metric. Profile link labels are copied from the article anchor text; they can be shortened, generic, or different from a profile's formal name. For identity matching, rely on the permalink.

Use with automated clients

The input and dataset fields support repeatable collection jobs. For example, a client can request “collect 40 recent venture articles and list their linked organization profiles,” then read the dataset's articleUrl and companyPermalinks fields. Another client can request “find coverage linking to this public organization profile during the last month,” passing that profile URL and a date window.

Treat every returned organization link as an article citation. If another system needs current profile details, it must obtain them from a source that actually supplies those details. Do not infer an amount, status, or contact from the presence of a link alone.

FAQ

Is it legal to collect this data?

The actor reads public news content. Your use of the resulting data should respect the site's terms, copyright rules, and applicable privacy law. Keep article attribution and source links when sharing summaries, and avoid using personal information in ways that the source does not support.

Do I need to set up a proxy?

No manual setup is needed for the default configuration. The actor uses its configured platform connection to request public news endpoints. If your account requires different connection settings, the proxy input is available.

How fast is a run?

It depends on the number of pages and filters. The 40-result validation run finished in roughly nine seconds of remote execution. A narrow organization filter can require more pages to find qualifying links, so it can take longer for the same number of delivered rows.

Can I schedule it and receive only new articles?

You can schedule it, but it does not store a prior-run baseline. Use articleId to remove duplicates in the receiving workflow. An overlapping date range helps catch articles that appeared close to a schedule boundary.

Can I export to a spreadsheet?

Yes. The dataset can be downloaded as CSV or JSON. Array fields can be expanded into a separate relationship table if a spreadsheet needs one line per article and linked organization.

Why is a field empty?

An article may lack a tag, image, byline, or linked profile. Taxonomy lookups can also be temporarily unavailable. Empty profile-link arrays mean no qualifying organization link was found in the article body; they do not prove that the article discusses no businesses.

Can I get company funding or headcount from a profile URL?

No. A profile URL is used only to filter news coverage and to identify links found in articles. This actor does not open profile pages or emit their funding, staff, investor, or contact fields.

Integrations

Run the actor on a schedule or invoke it from an external workflow, then read the default dataset through its URL. A data pipeline can store article rows keyed by articleId and expand the profile-link arrays into a related table. A notification flow can compare IDs with a prior run and send only newly published coverage. The source article URL remains the reference for every row.

Support

Open an issue on the Issues tab; we reply within 24h and add fields on request.