Protected Quick Tunnels: simple accountless authentication for your next dev project
- Source
- blog.cloudflare.com
- Date

Context
Cloudflare added email-based access control to Quick Tunnels, which expose a local development server through a temporary public URL. Developers can specify permitted addresses or domains without creating an account. Visitors prove control of an email address using a one-time PIN. Announcement
The access decision happens on the developer's machine: Cloudflare verifies the visitor's email, then the local connector checks that identity against its permitted list. The company says the list stays local. Protection requires the allowed-mail option; omitting it preserves the existing public-link behavior. Stopping the connector ends access, making this a way to share a temporary preview with selected people. Authentication flow
- Cloudflare Access only proves the visitor controls the email address; cloudflared makes the allow or deny decision locally against the rules typed on the command line, so the guest list never leaves the developer's machine.
- A stateless broker on Workers turns the Access identity into a short-lived signed assertion bound to the tunnel hostname and a single-use, 10-minute state, sent by form POST so it never lands in URLs, history or logs.
- Visitor sessions last up to four hours, or less if the Access sign-in expires sooner, and end when cloudflared stops. cloudflared strips auth credentials before forwarding, so the local app needs no login code.
- If the service does not confirm the authentication mode, cloudflared refuses to start instead of handing out a public URL, and a protected tunnel never falls back to public. Changing the means starting a new tunnel.
- Cloudflare suggests adding the flag to an instructions file such as AGENTS.md, then checking cloudflared's output, which reports whether email auth is on and how many rules it holds without printing the addresses.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- tool permissions — The rules governing which tools an agent may call and which need confirmation — the boundary between a mistake and an incident.
Checking sign-in…
Loading comments…

