Product updates
Trafficbeta

Cloudflare is now a Traffic provider

Collect AI referral visits, user fetches, and crawler events at the Cloudflare edge, on any stack, and let that evidence shape which Tasks come up first.

PromptScout Integrations page showing the Cloudflare Traffic provider with the Connect Cloudflare action and the three requested Cloudflare permissions.

What changed

  • Traffic now works on any stack Cloudflare proxies, not just a Next.js app on Vercel.
  • The collector runs ahead of your origin, so it records crawler hits on pages your application never renders.
  • What the connected website receives feeds weekly Task generation, so a gap on a page AI products already fetch outranks the same gap on a page nothing reaches.

Traffic now works on any stack

Traffic used to assume the collector lived inside a Next.js app on Vercel, which ruled it out for most websites. Cloudflare now sits beside Vercel as a full provider, and it is the better choice whenever Cloudflare already fronts the hostname visitors open. Your origin stops mattering: WordPress, Webflow, Shopify, a Rails app, a static bucket. If Cloudflare proxies it, the collector works.

Running ahead of the origin also changes what you can see. When GPTBot fetches a URL and your application never runs, a collector inside that application has nothing to report, while the edge collector already recorded it.

Three permissions, and what they rule out

PromptScout asks Cloudflare for three permissions and no others: Zone Read to identify the website you pick, Workers Scripts Write for one Worker and its secret, and Workers Routes Write for that Worker's route.

DNS is deliberately not among them, which is why an unproxied record is something you fix rather than something PromptScout fixes for you. The install stops rather than overwriting a Worker or route it does not own, and the installer authorization is cleared once setup finishes.

What it changes about your weekly Tasks

The connected website is not just another dashboard. Weekly Task generation now receives a bounded summary of what that site actually got: the observation window and total signals, the paths drawing the most attention, the AI providers behind them, and the referrers on attributed visits.

A citation gap on a page AI products already fetch is a different problem from the same gap on a page nothing has ever reached, and Tasks can finally tell those apart.

Availability

The guided one-click connection is in private beta while the public Cloudflare OAuth client finishes its release checks. Manual Worker installation is open to every eligible team today, and disconnecting removes only what PromptScout created.

Related

Cloudflare Traffic docs