Introducing Prometheus, an experimental Forward Deployed Agent for web data. Describe the web data you need and it writes Firecrawl code to collect it. Run it yourself or let us host and automatically maintain it as pages change. Try it with Claude Fable 5 for free this week!

Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. radar.cloudflare.com/traffic…

We've now fetched 8,000,000,000 pages at Firecrawl 🔥 A few other milestones in 2 short years: - 1.25M developers - 150K companies using us - 125K GitHub stars (top 100 repo) - 2.5M weekly downloads on npm PyPI. Thanks for building with us & we're just getting started!

We've now fetched 8,000,000,000 pages at Firecrawl 🔥 A few other milestones in 2 short years: - 1.25M developers - 150K companies using us - 125K GitHub stars (top 100 repo) - 2.5M weekly downloads on npm PyPI. Thanks for building with us & we're just getting started!

The web wasn't built for AI agents. So we built Firecrawl. - /search finds fresh, relevant sources from the web - /scrape turns any URL into clean context for AI - /interact automates clicks, pagination, and logins Join 1,000,000 developers connecting intelligence to the web.

The web wasn't built for AI agents. So we built Firecrawl. - /search finds fresh, relevant sources from the web - /scrape turns any URL into clean context for AI - /interact automates clicks, pagination, and logins Join 1,000,000 developers connecting intelligence to the web.

I dont love to gloat but we are almost 2x'ing openclaw just 3 days after surpassing their daily token volume 🤗

The Context Acquisition Company (CAC). We are a holding company acquiring services firms for their tokens in order to build domain specific agents/ models to deploy into our platform businesses and beyond. There's $1T hidden in the computers. We're gonna get it out.

One html page on avg produces 36k tokens. With proper md conversion you cut that down to 2.8k tokens. That’s 92% cost reduction While I agree that some cases HTML is better, by no means it’s a cost viable option for 99.9% of users You literally save $1,079* if u fetch 10k pages

Survivorship bias. For every unicorn, dozens of VC-backed startups burned billions extracting from LPs. I build for sustainable margins, not Series D valuations.

I love this idea from @jasonfried "Your only competition is your costs." Keep costs low, keep the team small, make stuff you want to use. You don't need the whole world: “A business is very simple. You got to make more than you spend. If you're making more than you spend, then your competition is your cost. That's what you're really in business against, how much it costs you to stay in business. It's not all the other alternatives that are on the market. You can't control what they're going to put out there, what they're going to price it at, all the things they're going to do. They're going to do what they're going to do. What I can control is how much it costs me to run my business, how much I sell my product for, and as long as I make more than I spend, I get to stay in business. And isn't that what this is all about, staying in business? That's what it's all about because I like this. I want to keep doing this. I can't keep doing it if I don't stay in business. I can't keep doing it if I make less than it costs me to make the things that I make. So I'm always thinking about the only competition I really have on an annual basis is to make sure that we make more as a company than it costs us to run the company. That's my real competition.”

Introducing Highlights for /scrape 🔥 Get the exact text, code, or tables you need from any web page using up to 100x fewer tokens. Just pass a URL and a query. Built for AI agents that require citable, hallucination-free context.

New Anthropic research: Natural Language Autoencoders. Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read. Here, we train Claude to translate its activations into human-readable text.