The same story gets published four hundred times. NewsCatcher (YC S22) collapses that noise into a single, structured event - and sells the clean version to the risk desks, banks, and AI agents that need it.
Open a browser during any busy news hour and pick a single event - a company gets acquired, a factory catches fire, a regulator issues a fine. Within a day, that one event exists as hundreds of near-identical articles, in a dozen languages, each with its own headline, its own timestamp, its own small inaccuracies. For a human skimming a feed, that redundancy is background noise. For a machine trying to understand the world, it is the whole problem. NewsCatcher, a company that came through Y Combinator's Summer 2022 batch, was built to solve exactly that unglamorous problem.
The pitch is a single sentence the founders keep returning to: turn the web into a structured database of real-world events. Not a feed. Not a search box that ranks ten links. A database - where one event is one row, deduplicated and enriched, no matter how many outlets covered it. That framing sounds modest until you try to build it, at which point it becomes a years-long exercise in scraping, cleaning, translating, and teaching software to recognize when two very different articles are describing the same thing.
Before there was a company, there was a Python package. Artem Bugara, a data engineer by training, published an open-source tool called newscatcher that let anyone programmatically collect normalized news from almost any website. People used it. People broke it. People asked for the things a free weekend project can never quite deliver - reliability, scale, coverage, support. That gap between what the open-source tool did and what serious users needed turned out to be the business.
Bugara studied econometrics at the Toulouse School of Economics and had run data engineering for one of the largest underwriters in aviation, energy, and space insurance - a world where knowing what happened, quickly and accurately, is the entire job. He teamed up with Maksym Sugonyaka, a self-taught data engineer with a data-science background, who became CTO. Both are Ukrainian; the company has roots in Kyiv and Paris before setting up a US base in Houston.
There is a pattern here that founders tend to learn the hard way and NewsCatcher stumbled into early. Give something away, watch what people actually do with it, and the paying product designs itself. The open-source repo was never going to be a company on its own. But it was a live feed of demand - every issue filed, every request for more sources or cleaner output, was a note on what the market would eventually pay for. The company grew out of that gap rather than out of a pitch deck.
"Turning the web into a structured database of real-world events."NewsCatcher's one-line description
The core product is the News API, now on its third major version. You send a query; you get back clean, deduplicated articles enriched with the things machines need - extracted entities (who and what the story is about), sentiment, and near real-time updates from that sprawling set of 90,000-plus sources. The v3 rebuild gave developers finer control over queries and richer data in each response.
Around that core sit two more products that reveal how the team thinks. The first is CatchAll, a web search API built on a contrarian idea. A normal search engine ranks results and shows you the best ten. CatchAll does the opposite: for a question whose answer is spread across hundreds or thousands of pages - list every regulatory action this quarter, every funding round, every product recall - it retrieves a large candidate set, validates the pages, and hands back structured records. When your question is a long list, ranking is the enemy and recall is the product.
The second is the Local News API, which indexes hyper-local US coverage across more than 31,000 locations. Local newsrooms have been shrinking for two decades, which is precisely what makes their output scarce, scattered, and valuable to anyone trying to monitor what is happening in a specific place.
It is tempting to assume a news company sells to people who like reading news. NewsCatcher's customer list points the other way. Its public references include the US Department of State and Samsung, and its stated focus is on enterprise risk teams, financial services, government bodies, media intelligence firms, and - increasingly - AI platforms. These are organizations that treat news as a signal, not a story: an early warning about a supply chain, a sanction, a competitor, a crisis.
A news article isn't the unit that matters. The event is. Ten thousand articles about one earnings call are one row in the table you actually want.
The distinction matters because the buyers behave differently from readers. A reader forgives a missed story; a risk desk does not. A reader tolerates ten versions of the same headline; an analytics pipeline chokes on them. When your customer is a bank monitoring exposure, a government body tracking events abroad, or a media-intelligence firm selling coverage reports, the qualities that matter are completeness, cleanliness, and speed - not editorial voice. NewsCatcher's coverage stretching across many languages, with recent gains in Chinese and Japanese article volume, exists because those regional signals are exactly what supply-chain and financial teams cannot afford to miss.
That reframing explains almost every product decision. It is why deduplication is treated as a feature rather than a footnote. It is why the company pursued SOC2 Type II compliance and ISO certification - enterprise and government buyers do not sign until the paperwork does. And it is why the roadmap has tilted toward AI: an agent that needs to read the web needs something underneath it that has already done the collecting, cleaning, and de-noising. NewsCatcher spent years building that plumbing before agents were fashionable.
Type "news API" into a search bar and you get a long list of providers - general-purpose news feeds, web search APIs, open datasets like GDELT, scraping platforms, and cloud vendors' own search products. On the surface they look interchangeable. The difference NewsCatcher argues for is not access to articles, which is close to a commodity, but the absence of the mess that comes with them: duplicates removed, entities resolved, sentiment attached, languages normalized, and recall prioritized where recall matters.
The model is business-to-business, data-as-a-service: subscription and usage-based access to the News, Web Search, and Local News APIs. The funnel began with the free open-source package, which did the cheapest and most honest product research there is - it showed the founders exactly what people tried to build and where the free version fell apart. In late 2022 the company raised a $1.5M seed round, backed by Y Combinator, TinySeed, and Uncommon Capital.
A free tier that funnels into paid infrastructure is a familiar developer-tools playbook, but it works only when the paid part is genuinely hard to replicate. Deduplication at scale, multilingual normalization, entity resolution that keeps two different "Apple" companies apart, and the operational grind of keeping tens of thousands of sources flowing are not weekend problems. That difficulty is the moat. A customer could, in theory, scrape the same sites themselves. In practice, the cost of building and maintaining the cleaning layer is exactly what they are paying to avoid.
The bet underneath it is that the most valuable APIs are invisible. Nobody posts about how much they love their news-deduplication layer. But when that layer breaks, a risk dashboard goes dark and a monitoring system starts missing events. NewsCatcher is choosing to be infrastructure rather than a brand - the boring middle of the stack that everything above it quietly depends on.
Put NewsCatcher on a map of the software world and it lands in an odd, useful spot: between the chaos of the open web and the neat dashboards, models, and agents that want to consume it. Its job is translation - taking something built for humans to skim and turning it into something a machine can query. As more products lean on live information to make decisions, the demand for a clean reading layer grows, and the companies that built the plumbing early tend to be the ones still standing when the hype settles.
There is also something quietly countercyclical about the whole enterprise. As local newsrooms close and the open web fills with duplicated, machine-generated, and lightly rewritten copy, the raw material NewsCatcher works with gets messier every year. That could look like a headwind. Framed differently, it is the reason the company exists at all - the harder it becomes to tell one real event from four hundred echoes of it, the more valuable a service that does the untangling becomes. The mess is not a bug in the market; it is the market.
For now, NewsCatcher's ambition remains legible in that one repeated sentence. Somewhere out there, the same event is being published for the four-hundredth time. The company's whole reason for existing is to make sure that, to the systems that matter, it still counts as one.