Hey, good morning everyone. Thanks so much for joining us today. I'm so excited for what we're about to share. So for a decade now, code running on Vercel was primarily built for immediate web requests. But the most interesting software our customers create today no longer fits that model. And the number of these new projects is exploding just in the last few months.
Weekly deployments in Vercel have doubled, and a third of those new deployments come from coding agents, a number that's gone up 1,000% in just six months. But the most interesting part to me isn't how much we're deploying, it's what we're deploying. Those agents are shipping agents and the agents that our customers build. They don't just respond once and wait. They need to reason, call tools, run humans, and then do it all again.
And each of those workflows needs a backend that can keep up. So what I've seen in Vercel is that the teams shipping the fastest right now, they aren't rebuilding infrastructure and coding queues themselves, or writing retry logic from scratch, or even wiring up observability. The stuff's been done. So they're using our infrastructure primitives, the stuff we've spent the last decade building so they can focus all of their focus on their product. And that's exactly the problem we built workflows to solve.
We launched it in beta six months ago, and the response was incredible. And now, 100 million runs later, we're so excited to announce that workflows is GA today. Let me show you how it works. The app I'm going to show you is a birthday card generator that I built. Now it could be anything really, but this hits all the things we just talked about. It doesn't fit into a single request. It waits on humans, it retries when things fail, and it streams progress in real time to the user. So here's how it works.
Let me send myself a card, and I'll start by entering my email. Next, I'll pick a date. And I'm going to pick the 23rd of April. Now I like doing bonfires on my birthday, so I'll do a beach bonfire and I'm sending it to myself. So for RSVP, I'll enter my own email here just to make the demo easy. So it's going to have myself RSVP to my own birthday. Cool. Now I'm going to hit Generate and Send and I'll see the birthday card start generating.
I'll head over to the Vercel dashboard where I can see this run just came in. Let me click in. And I can see this in real time as it generates prompts and then generates the message and generates an image as well. This is the observability you were talking about. Every step is visible right here. You can see exactly what's running, what's completed and what's waiting.
Over on the user side, I can see that the card's been generated and the RSVP emails have been sent. At this point, the workflow is suspended. There's no compute actually happening. It's waiting for a human to respond. And I could go into my email, respond right now, or even wait for a few hours or days and respond. Then this will continue to track the progress.
Now, before I show you the email, I actually want to show you the CLI workflows ships with. The CLI on day one that lets you inspect everything that's available on the observability dashboard. This means that you and your agents have access to the complete picture when you're trying to debug something that went wrong. Okay, over to my email. I should see the email come right in. There it is. And it's just asking me to RSVP. I'll hit yes, that'll be there.
Back to the dashboard. I can see the hook's resumed and it should kick off into a sleep. Back to the user dashboard and I'll see the same thing. We're no longer waiting for RSVP, and now it's just waiting for the 23rd so I can send out the birthday. Now I don't want to make you wait for two weeks, so I'll head back to the dashboard and wake up the sleep right here. That should immediately resume the workflow, and we'll see the workflow complete when we head back to the user dashboard. Postcard's been sent and over in my email I should get a birthday card with the RSVP as well.
Cool. Now I want to show you the code and remember the thing we said earlier. Teams shipping the fastest today aren't dealing with queues and managing state and infrastructure. That's the old way. In workflows, everything we've seen today is a single function and it starts with this directive: use workflow. This one directive is what lets this function run for days, suspend between steps and resume exactly where it left off. So when the workflow calls a step like generate prompts or generate image, the workflow suspends the function, queues the work in the background, retries if it fails, and then resumes the workflow when it's done.
The entire time, I get full observability on the Vercel dashboard of every step along the way. I can even click into one of these and by default on Vercel, everything's encrypted, but I'll decrypt the data so you can see the input. Yeah, this is available both on the dashboard and the CLI. Workflows also come with built-in streaming. This is how I push real-time updates to the client to track progress live. I didn't have to pull in some sort of database just to pull this off. The client can even disconnect and reconnect just fine. This is a persistent stream.
All right. And here's the hook. For every RSVP, I'm creating an ephemeral webhook URL on the fly. I'm then sending an email, and then I can simply just wait for it down here. The workflow pauses right here until every person clicks on their link. There was no database involved, and the results even just live in local JavaScript variables. The workflow holds all of its own state. This is the distributed systems machinery that we said was just gone. You don't have to think about this stuff anymore.
All right, lastly, let's look at the sleep at the end to sleep for my birthday. It's right here. The sleep could be a couple of seconds, days or even weeks like we just did. And while it's sleeping, the workflow is just suspended with zero compute running. When it wakes up, it continues exactly where it left off, maintaining state along the way. And this is the entire thing. It's one function that you can just read top to bottom and reason about from end to end. Every step is backed by a queue, retries automatically and runs on its own serverless function. There are no timeout limits, so workflows can run as long as they need to. This is what we mean when we said Vercel handles everything underneath. You can just write the logic, so we take care of the rest.
Now in this demo I showed you a simple use case, but our customers use workflows to run agents in production at tremendous scale. Mocks uses workflows to ensure video generation pipelines run reliably every time. And Flora is orchestrating parallel creative agents across 50 image generation models all with workflows. Durable is a small team that hosts 3 million customer websites, and they use workflows to ship new agents to their users every single day.
Now I started this demo by saying that the most interesting software our customers build no longer fits in a single request, and we're seeing more and more apps like these deployed every single day. Agents are shipping agents, and the teams building them aren't thinking about queues because they're focusing on their customer problems. And the shift isn't just happening inside Vercel. Our customer deployments are doubling and fewer engineers than ever are hand-rolling infrastructure. So whatever app or agent you want to build, workflows will make it reliable, long-running, durable, and observable. From your laptop to a million users, all with the same application code. Workflows is generally available today, so go achieve something that works.
