Ship to
the edge.
Everywhere at once.
Relay serves and runs your code from 340+ cities in 128 countries. Static, dynamic, and streaming — answered from the point of presence nearest your visitor, in single-digit milliseconds.
Measured from where
your users actually are.
Live p95 latency sampled from real client requests over the last 60 seconds. No single-region fallbacks, no cold origins — every edge holds a warm copy.
One network.
Every layer of the request.
CDN, compute, security, and DNS answered from the same node — so a request never leaves the edge to be resolved, cached, filtered, or rendered.
Smart tiered caching
Content is pinned across regional shields, so origin fetches drop to near zero. Tag-based purge propagates worldwide in under 150ms.
98.7% cache hit ratioEdge Functions
Deploy JavaScript, TypeScript, and WASM to every POP. V8 isolates boot in under 5ms with zero cold starts — no containers, no regions to pick.
<5ms isolate bootAlways-on DDoS shield
248 Tbps of absorption capacity mitigates L3–L7 floods at the edge before they reach your origin. Rules update globally in seconds.
248 Tbps capacityEdge KV & cache API
Strongly-consistent key-value storage replicated to every region. Read your data from the same isolate handling the request — no round trip.
<1ms local readsAnycast & instant DNS
A single anycast address routes each visitor to the closest healthy node. Authoritative DNS resolves in a median of 9ms worldwide.
9ms median resolveReal-time analytics
Per-request logs, latency histograms, and cache diagnostics stream to your dashboard with sub-second lag. Query 30 days of edge traffic.
<1s log latency// runs in every one of 340+ cities export default async function handler(req) { const geo = req.cf.geo; // resolved at the edge const hit = await Relay.KV.get(geo.city); if (hit) return new Response(hit, { headers: { "x-relay-pop": geo.pop } }); const data = await render(geo); // <5ms isolate Relay.KV.put(geo.city, data, { ttl: 3600 }); return new Response(data); }
Push once. Run in every city.
One relay deploy ships your function to the entire fleet. No region selection, no autoscaling config, no idle instances — the code lives where your users are and scales to the request.
- Cold start
- 0 ms · always warm
- Deploy time
- Global in 4.1s
- Runtime
- V8 · WASM · TS
- Free requests
- 1M / day
From click to
byte in four hops.
Every request rides the same optimized path — resolved, routed, filtered, and answered without ever leaving Relay's backbone.
Anycast resolve
DNS points your visitor to the single nearest POP via one anycast IP — 9ms median, worldwide.
Shield & filter
L3–L7 mitigation and WAF rules run inline at the edge, dropping hostile traffic before it costs you.
Cache or compute
A warm cache answers instantly; a miss spins a V8 isolate in under 5ms and stores the result nearby.
Stream back
The response streams over HTTP/3 from the same node — no backhaul, no origin round trip.
Priced by traffic,
not by region.
Bandwidth is metered flat, worldwide. No per-region multipliers, no egress traps — the same rate whether your users are in Ohio or Osaka.
- 1M edge requests / day
- 100 GB egress included
- Global CDN + auto TLS
- 3 Edge Functions
- Community support
- 100M requests / day
- 2 TB egress, then $0.04/GB
- Unlimited Edge Functions
- Edge KV + real-time logs
- DDoS shield · 99.99% SLA
- Committed-use bandwidth
- Dedicated POP capacity
- Custom WAF & bot rules
- Private backbone peering
- 24/7 support · named SRE
The edge is
already there.
Point your domain, push your code, and let Relay answer every request from the city nearest your visitor. First deploy takes about a minute.