A New Tool That Lets AI Agents Write Their Own Playbooks

Here's what's actually interesting about what Cloudflare just shipped: it's not just another developer tool. Dynamic Workflows is a TypeScript library built on top of Cloudflare's existing Workflows engine and its recently beta-launched Dynamic Workers — and what it does is kind of remarkable. It lets each step in a workflow survive failures, sleep for hours or even days, wait for external events, and then pick back up right where it left off. Even if the underlying isolate gets recycled.

That last part matters more than it sounds. Long-running AI jobs fail all the time because the environment they're running in disappears mid-task. Dynamic Workflows is designed specifically to solve that. And crucially, tenant-provided code can be routed on the fly — meaning developers don't have to register every workflow in advance. The system figures it out.

Cloudflare put it plainly in their announcement: the vision here is that "the agent writes the workflow; the platform runs it; neither has to know ahead of time what the plan looks like." That's a pretty bold way to frame it, but honestly? It captures exactly what makes this different.

Built for Agentic Workloads at Real Scale

What the Numbers Actually Look Like

This isn't a small experiment. The Workflows engine now supports up to 50,000 concurrent instances and 300 new instances per second per account. Cloudflare says the redesign was made specifically with agentic workloads in mind — which tracks, given how much infrastructure AI agents actually demand when they're doing anything remotely complex.

Think about what an AI agent actually needs to do: pause mid-task, wait for a human to respond, retry a failed API call, hand off to another agent, sleep overnight. Most traditional infrastructure wasn't built for that kind of stop-and-go behavior. Dynamic Workflows is.

The Earnings Context You Probably Shouldn't Ignore

Cloudflare Reports Q1 2026 on May 7

The timing of this launch is worth paying attention to. The announcement came just days before Cloudflare is set to report its first-quarter 2026 earnings after market close on May 7 — alongside an Investor Day. That's not a coincidence. Product launches like this signal to the market where growth is coming from.

And the market noticed. Cloudflare shares climbed to around $217.50 on May 1, up from a close near $205 the day before. That's roughly 11% up year to date, with investors clearly positioning ahead of the earnings report.

What Analysts Are Expecting

The company's own guidance points to Q1 revenue of $620–$621 million, which already beat prior consensus of $615.50 million. Analyst consensus has since drifted up to around $622 million — that would represent approximately 30% year-over-year growth if it lands.

Baird analyst Shrenik Kothari upgraded Cloudflare to Outperform back in February with a $260 price target, calling Q4 2025 results the "strongest confirmation yet" of expanding growth drivers across AI infrastructure, Zero Trust security, and the Workers developer platform. TD Cowen and RBC Capital have both maintained bullish stances with $265 price targets.

The Real Question Heading Into May 7

Can Product Momentum Hold Up the Premium Valuation?

Here's the honest tension in all of this. According to Simply Wall St, Cloudflare's stock is trading more than 119% above one discounted cash flow estimate. That's a steep premium by any measure — and it means the earnings report on May 7 isn't just about revenue. It's about whether the company's push into AI developer tools and agentic infrastructure can keep justifying investor enthusiasm.

The Investor Day happening alongside the results will probably matter just as much as the numbers. Cloudflare has been methodically building out a developer ecosystem — from Workers to Dynamic Workflows — and Investor Day is where that story gets told at scale.