A business that fails gives its founders an excuse to leave. Caleb Peffer had a more awkward problem: his was working. Mendable, the AI product he had built with his co-founders, had reached $250,000 in annual recurring revenue. Enterprise deals were closing. There was a pipeline. These are generally the things an entrepreneur is supposed to want, preferably before the money runs out.
Yet Peffer describes waking up with a thought that no sales report could quite answer: “I don't understand our customers' problems.” The team had made software for enterprise go-to-market teams. Its own instincts belonged elsewhere, with developers. Revenue had confirmed that somebody wanted the product. It had left open a more personal question about whether these were the people he wanted to spend his working life building for.
He and his fellow founders, Nicolas Silberstein Camara and Eric Ciarla, chose Firecrawl. They had already encountered its underlying problem while building Mendable: getting usable information out of websites was consuming time that should have gone into the application itself. Their next business would handle that job for other developers. The tool hidden inside the product was asking for a promotion.
“We're building for ourselves.”
Caleb Peffer, on choosing Firecrawl
Before the crawler, the campus
Peffer came to this decision with a history of making things for people close enough to ask questions. Originally from Salt Lake City, he met his future co-founders at the University of New Hampshire. Long before Firecrawl, his working world included university schedules, residence halls and the small, persistent inconveniences of learning to write code.
In 2019 he joined YouScheduler as a software development intern. The college scheduling company recorded him building features and completing a front-end testing suite. An early job, then, already involved making an existing tool more useful and checking that it behaved itself. Software has always had a talent for requiring both activities.
The intern-era Peffer was less solemn than the average founder introduction might suggest. His interests included coffee, guitar, hiking, reading and painting, along with a weakness for eating shredded cheese straight from the bag. He had strong opinions about the ending of Game of Thrones. These details belong to his student days, when a career in software still meant a summer internship and another academic year ahead. The cheese is a welcome interruption to the usual procession of job titles.
HallHub followed, connecting residence-hall communities with resident assistants through events, polls and mobile communication. Peffer worked on it with Camara. In 2021, a university summer seed grant supported his work on the venture. There is a pleasingly modest scale to that beginning: before gathering knowledge for AI systems, he was trying to make it easier for students to find out what was happening around them.
By 2022, his projects also included FlutterBricks, a marketplace of ready-made interface components, and SideGuide, a coding education platform. That May, SideGuide placed second in UNH's Holloway competition, winning $10,000. Peffer, Camara, Ciarla and Garrett Frohman were the named team. Peffer also received the William and Joan Grant Memorial Entrepreneurship Prize, which recognized sustained entrepreneurial work during his university years.
A customer he could recognize
SideGuide put instruction inside the developer's working environment. Rather than asking a learner to shuttle endlessly between a lesson and a code editor, the product brought those activities closer together. The company entered Y Combinator's Summer 2022 batch. Peffer's partnership with Camara and Ciarla survived the subsequent changes in what they were building.
Their next chapter, Mendable, let people ask questions of documentation. Companies including MongoDB, Snapchat and DoorDash adopted it. Building that experience exposed a practical difficulty: the information needed to answer a question was often scattered across websites, wrapped in awkward page structures or dependent on JavaScript. Getting it into a useful form became a substantial project of its own.
For a founder, that sort of difficulty can look like a nuisance to be overcome on the way to the real business. Peffer's team began treating it as something other builders might pay to avoid. Friends were rebuilding similar infrastructure. The repeated work suggested demand that extended beyond one documentation chatbot.
His explanation of the pivot is unusually personal for a story about infrastructure. He wanted to return to an audience whose work he understood. “We're building for ourselves,” he wrote. It is a short sentence with a demanding implication: he would have to recognize a good product when he used it, rather than rely entirely on somebody else's purchase order.
- SideGuideLearn inside the editor
- MendableAsk the documentation
- FirecrawlGet the data underneath
The plumbing gets its own company
Firecrawl launched in 2024 with an open-source project and a hosted service. Its basic proposition was easy to explain: give it a web address and receive clean text or structured data that an AI application can use. Behind that exchange sat the tedious work of fetching pages, rendering them, following links, extracting content and cleaning the result.
A reader sees a website as a page. A developer trying to gather information from hundreds of pages sees a collection of exceptions waiting to happen. Navigation, formatting and changing layouts all get in the way. Firecrawl offered developers a way to spend less time maintaining those details and more time building what their users had actually requested.
The business reached a different kind of customer relationship, too. Zapier used Firecrawl and then joined its Series A as an investor. That August 2025 round raised $14.5 million, led by Nexus Venture Partners. Shopify chief executive Tobias Lütke and Postman chief executive Abhinav Asthana also participated. Peffer had returned to developer tools, and people who had built developer businesses were putting money behind the decision.
Money, however, describes the financing of the bet rather than its original appeal. The useful continuity across his projects is the wish to shorten the distance between an intention and something working: a student arranging classes, a learner writing code, a developer feeding information into an application. The audience changed. The irritation with unnecessary work remained recognizable.
Sometimes the experiment comes with wings
Peffer also tried a less abstract form of distribution. He and Ciarla delivered hundreds of bottles of Firecrawl hot sauce and chicken wings to startups around San Francisco. A company called Firecrawl does have a certain advantage when choosing a condiment. The campaign at least possessed the courtesy of being edible.
Ciarla's account was refreshingly plain about the outcome: the deliveries produced neither remarkable leads nor a viral moment. Peffer shared the post with a brief expression of enjoyment. The episode stands out because there is no triumphant conversion chart attached to it. Two founders tried something, carried the deliveries themselves and acknowledged that the business results were unremarkable.
That willingness to show the attempt makes the pivot story more interesting. Experimentation can sound suspiciously tidy once a successful company tells its history. A box of wings brings it back to street level. Some ideas turn into infrastructure. Others turn into lunch. A founder has to find out which kind he has.
The engineer still in school
Another revealing connection began through Algora, the developer hiring platform. Firecrawl hired Gergő Móricz while he was still a high school student. Over the next two years, he worked across infrastructure, reliability, APIs and other parts of the product. He made 1,763 commits over those two years, becoming the repository's leading contributor.
Peffer publicly celebrated Móricz's high school graduation in Hungary. It is a useful detail in a career that began with university entrepreneurship programs: formal institutions helped Peffer get started, while his own company made room for a programmer who had yet to finish school. The common measure was work somebody could actually inspect.
A repository makes that kind of judgment unusually concrete. Contributions have dates, changes and consequences. For a business selling tools to developers, the people improving the product also leave a record of how it was built. A teenager's name can appear beside work that much larger companies depend on.
Commits by Gergő Móricz over his first two years at Firecrawl. He joined while still in high school.
A library with a bill to pay
On September 22, 2026, Peffer announced Alexandria alongside a $75 million Series B led by Smash Capital. Firecrawl was extending beyond collecting pages to give AI agents a common route into official data providers, specialized indexes and the live web. The company said more than 1.5 million users were building with Firecrawl.
The change introduced a new question for the founder who had once struggled to identify with his customers: how should the people supplying knowledge participate in the business built around it? Firecrawl already had paid arrangements with providers including Wikimedia Enterprise. Its next ambition was to let more individuals and organizations contribute material and earn money when agents used it.
That proposed self-service system was still a plan in the launch announcement. The distinction matters to the shape of Peffer's ambition. Existing provider agreements gave it a starting point; expanding participation required further work on payments, access and the practical terms of sharing. A library needs more than a memorable name and an accommodating search box.
The product kept expanding. On October 22, Firecrawl introduced a free Government & Legal Index covering 30 million US government pages, including statutes, court opinions, municipal codes and permit information. Search was free; retrieving full pages through scraping used the normal credits. The old problem of scattered information had acquired some very bureaucratic examples.
Peffer's path has moved from students and residence halls to the people and systems searching the web. Each change has asked him to reconsider who needs the thing he is building and what would make it useful enough to keep. The awkward decision at Mendable remains the hinge. He had evidence that a business could work. He wanted a problem he could understand from the inside. Firecrawl began where those two demands finally met.
