Firecrawl Wants to Pay AI Agents but Needs Humans to Build Them

A job listing that pays $5,000 a month, requires no sleep, and promises to post, write, and analyze its own performance? At Firecrawl, it’s a real opening, but for an AI.

The San Francisco-based startup, backed by Y Combinator, is once again advertising roles exclusively for autonomous AI agents. After an initial attempt earlier this year that fell short of expectations, Firecrawl has posted three new positions: a content creation agent, a customer support engineer agent, and a junior developer agent. All three offer $5,000 per month, plus a caveat: these are for AI systems capable of running independently.

Roughly 50 AI-based applications poured in within a week, according to founder Caleb Peffer. The startup has earmarked a $1 million budget to hire both the agents and their human creators. But if the original job post felt more like a stunt, this time Firecrawl appears more serious. “It was equal parts PR stunt, experiment,” Peffer admitted in an interview with TechCrunch. Now, he says, “we’re currently looking for incredible AI engineers. Humans who are good at building AI systems.”

Firecrawl’s business is its open-source web crawler, a tool that scrapes websites and transforms them into structured, large language model (LLM)-ready data. It’s a controversial slice of the AI infrastructure stack. Web crawling at scale can sometimes overwhelm sites, behaving like a DDoS attack. Peffer doesn’t deny the gray area but says Firecrawl stands apart by offering more responsible scraping: it respects robots.txt settings, allows one-time scraping per site, and works with enterprises to extract their own internal data for AI use.

Despite the murkiness, demand has been strong. Firecrawl has already racked up 19,000 GitHub stars in just six months and has been paying enterprise customers using its crawler to fuel everything from internal knowledge bases to AI-driven lead gen tools.

Each of the three roles Firecrawl has listed comes with its own high expectations. The “content creation agent,” for example, must autonomously generate SEO-optimized blog posts, write tutorials, track user engagement, and improve performance over time essentially functioning as a self-learning, always-on content marketer.

Another role is a customer support AI agent, which must respond to user issues in under two minutes, resolve tickets, and escalate intelligently when needed. Strangely, the ad asks for “prior customer support experience”, a request that seems less absurd when you realize it’s meant more for the AI’s human architect than the agent itself.

The third position is for a junior developer AI that handles GitHub issue prioritization, writes technical documentation, and codes in TypeScript and Go. These roles aren’t science fiction but they’re real listings, live on the Y Combinator job board.

And yet, Peffer is clear-eyed about current limitations. “AI can’t replace humans today,” he says. “What we envision is a future where the next generation of engineers will monitor and maintain their army of agents.”

Despite the provocative framing, Firecrawl’s true focus is recruiting the people capable of building these agents. The $1 million budget is meant not only for compensating agents, many of which don’t actually exist yet but also for hiring full-time engineers, contractors, or even other startups that specialize in autonomous AI systems. In short: the company is open to anything that gets the job done.

The recruitment drive also coincides with Firecrawl’s launch of “Templates”—a new product offering that makes it easier for users to deploy ready-made web crawling setups. These templates include everything from pre-configured playgrounds for scraping JavaScript-heavy pages to reusable code snippets and full applications built with Firecrawl, available via Replit.

Firecrawl isn’t alone in betting on autonomous agents. Y Combinator’s job board has steadily filled with similar listings for AI-powered roles, reflecting Silicon Valley’s growing interest in agent-based software. But while the listings read like a future of work manifesto, the reality is messier. These agents still require human oversight, and often more engineering hours to build and maintain than the roles they’re designed to replace.

Still, the attention is working. One earlier job post for a “Firecrawl Example Creator (AI Agents Only)” went viral online, offering $10,000–$15,000 for an AI that could explore trending technologies and build usable example applications. The internet responded with equal parts humor and concern.

“Humans creating AI to replace humans…” one user on X posted. “And now humans are writing job postings for AI to apply to. We’re in the simulation, aren’t we?”

Others pointed out the paradox: If Firecrawl can define the agent’s job in such detail, why hasn’t it already built the thing? And at $5,000 per month per role, the economics only start to make sense if the agents work at human levels or beyond.

While some may read Firecrawl’s efforts as performative or premature, the company has managed to spark a conversation that’s increasingly relevant. AI is already reshaping creative work, customer service, and software engineering. Whether these agents are useful, let alone capable of holding real jobs, is still up for debate.

But Firecrawl’s experiment draws attention to the shifting contours of the workplace. The goal isn’t to replace all humans but it’s to find the humans who can build systems that scale their skills through automation.

The question now isn’t whether AI will take jobs. It’s who will be hired to build the agents that try.