The 37-person startup pipes real-time company and people data into AI agents. Seven months out of Y Combinator, revenue jumped from $770K to more than $4M - with one salesperson and no ad spend.
Almost every startup pitching investors in 2025 sold some version of the same dream: an AI agent that would sell for you, recruit for you, or vet a deal for you while you slept. Crustdata looked at that stampede and asked a quieter question - what, exactly, is the agent supposed to know? An agent that cannot see who just changed jobs, which company just raised money, or where headcount is climbing is a very confident tool working from a very old map.
Crustdata, a Y Combinator Fall 2024 company based in San Francisco, sells the map. It provides live company and people data through clean APIs, webhooks, and full dataset delivery, pulling from more than a dozen primary sources and keeping it fresh enough that an AI agent can act on it in real time. The founders describe the mission plainly: make hard-to-get data easy to use at scale.
The traditional B2B data business was built for humans - analysts clicking through dashboards, exporting spreadsheets, filtering lists. Crustdata starts from a different assumption: the customer is an AI agent making an API call, and it needs an answer it can use in the same second. That single design choice runs through the whole product line.
There is a People and Company Search API for finding and enriching profiles, a Web Search API that lets agents pull real-time web signals, and Bulk Datasets for teams that want everything at once - real-time headcount, people changes, funding, website traffic, product reviews. The piece customers tend to describe first, though, is Watcher.
Watcher turns a static database into a stream of events. You subscribe to a set of companies or people, and Crustdata fires a webhook the moment something happens that you told it to care about.
For a sales team, that is the difference between calling a prospect the week they got promoted and calling them six months late. For a recruiter, it is knowing an engineer is on the move before the job board does. For an investor, it is a funding round surfacing in a feed instead of a news alert everyone else already read.
The technical claim behind all of this is that the data comes from primary sources - more than fifteen of them - rather than a licensed copy of someone else's aging database. Crustdata builds dedicated pipelines that pull directly, so the numbers it reports for headcount, people changes, funding, web traffic, and product reviews reflect what is happening now rather than what a quarterly refresh captured. Freshness is not a feature here so much as the entire proposition; a stale signal fired late is worse than no signal at all, because it teaches an agent to act on the wrong moment.
Crustdata's users are the AI-native companies building on top of this shift: AI SDRs, sales copilots, sourcing agents, and due-diligence tools. Three named customers give a sense of the range. Y Combinator uses Crustdata to help discover and recruit founders - the accelerator is both an investor and a paying user. Dharmesh Shah's agent.ai, from the HubSpot co-founder, taps its data to power agent workflows. And MNTN, the performance-marketing company associated with Ryan Reynolds, uses it for AI-driven sales.
The through-line is that none of them wanted to build a data pipeline. They wanted to build an agent, and they needed the agent to be right about the world. Crustdata is the part they chose not to build themselves.
That is a familiar pattern in infrastructure. The teams closest to a hard problem are usually the ones best positioned to sell the solution, because they have already felt the pain of the alternative. Every one of these customers could, in theory, assemble its own data layer from a mix of scrapers and third-party vendors. The reason they do not is the same reason most companies do not run their own power plants: the work never ends, it is nobody's core product, and a specialist doing it full time will simply do it better. Crustdata is betting that the maintenance burden of live B2B data is heavy enough to keep that logic intact as the agent market grows.
The number that made investors sit up was not the size of the round. It was how the revenue got there. Crustdata reports scaling from about $770,000 to more than $4 million in roughly seven months after leaving Y Combinator - with zero paid marketing and a sales team of one. That is the profile of a product people go looking for, not one that has to be sold to them.
It is also a legible signal to a certain kind of buyer. When developers adopt a data API on their own, kick the tires, and expand usage without a sales rep in the room, it usually means the alternative - scraping sources by hand, stitching together three vendors, babysitting stale records - was painful enough that a clean API felt like relief.
Growth like that is rarely an accident of timing alone. It tends to follow when a product lands in the exact window where demand is forming faster than supply. AI agents went from demos to production budgets over the same stretch, and each new agent arrived with the same unmet requirement for live data. Crustdata did not have to create that demand; it had to be the cleanest option in front of teams already looking. The single-salesperson detail is less a growth-hacking flex than evidence that the market was pulling the product forward rather than the company pushing it out.
Crustdata's founders did not arrive at data by accident. Two of them met through Y Combinator's co-founder matching tool in 2023, and the third joined to round out the team. Between them they had already built and sold data products and run a private-company database used by large enterprises.
The résumés matter here because the hard part of this business is not the API - it is the plumbing behind it. Pulling from many primary sources, reconciling them, and keeping the result fresh is the sort of unglamorous data engineering that Chowdhary spent a decade doing at scale before it became a startup.
B2B data is not a new category. ZoomInfo, Apollo, People Data Labs, Clearbit, and PrivCo have sold company and contact data for years, and enrichment layers like Clay sit on top of them. Crustdata's argument is not that those tools are wrong - it is that they were built for a human clicking through a dashboard, updated on a slower clock, and awkward to wire directly into an autonomous agent.
| Dimension | Crustdata | Traditional B2B data |
|---|---|---|
| Built for | AI agents / API calls | Human dashboards |
| Freshness | Real-time, event-driven | Periodic refresh |
| Change alerts | Watcher webhooks | Manual list re-pulls |
| Primary sources | 15+ pulled live | Varies, often licensed |
| Delivery | APIs + bulk datasets | Exports & UI |
Whether that gap holds as incumbents add their own agent features is the open question. But the timing works in Crustdata's favor: it built for machine consumers from day one, while the rest of the category has to retrofit.
In October 2025, Crustdata closed a $6 million seed round led by Y Combinator and A Capital, with General Catalyst, SV Angel, and Phosphor Capital participating. Notably, Garry Tan doubled down - the accelerator that first backed the company, and that also uses its product, put more in as lead. A customer who becomes a repeat investor is about as direct a reference as a data company can get.
The business model underneath is a credit-based API subscription. Reports put entry pricing around $200 a month for individual and starter use, scaling to custom plans for platforms and enterprises that consume data in volume. It is the kind of usage-aligned pricing that rewards exactly the product-led adoption the company has leaned on.
Zoom out and Crustdata sits at a useful intersection: the AI infrastructure layer and the decades-old B2B data market, meeting at the exact moment agents need something to read. If the next wave of software really is autonomous agents doing the work of sales reps, sourcers, and analysts, then a clean, live data feed stops being a nice-to-have and becomes a dependency. The companies that win the picks-and-shovels position in a boom rarely make the loudest headlines, but they tend to make the most durable businesses.
There is a risk in that position worth naming. Infrastructure companies live and die on trust and continuity - customers wiring an API into their core product need it to be accurate, available, and stable for years. The upside of being the quiet dependency is that switching costs run in your favor once an agent is built on top of you; the downside is that any slip in data quality is felt immediately and everywhere. Crustdata's founder résumés, heavy on large-scale data engineering, read like an answer to exactly that concern.
Crustdata is not promising to be the smartest agent in the room. It is trying to be the thing every smart agent quietly calls. For a 37-person company two years past a co-founder match, that is a specific and defensible place to stand - and, if the agent wave arrives the way its customers are betting it will, a valuable one.