compute-edge_
Edge runtime · Tallinn, EE

Run code where your users are. No regions.

A globally-distributed runtime for low-latency apps. Write a function in TypeScript, Rust, or Python — get a URL that resolves at every PoP. We do the routing, you do the work.

Read the docs →
▸ Deploy in 14 seconds

One command. Every PoP. No infrastructure to babysit.

~/hello-world — compute-edge
$ compute-edge deploy
  reading compute-edge.toml...
  building hello-world (v0.4.2)...
  ↳ 312 KB bundle, 4 routes
 
  deploying to 14 edge PoPs:
  ✓adl-01 · Adelaide, AU284 ms
  ✓syd-02 · Sydney, AU190 ms
  ✓per-01 · Perth, AU312 ms
  ✓akl-01 · Auckland, NZ218 ms
  ✓sin-01 · Singapore340 ms
  ✓nrt-02 · Tokyo408 ms
  ✓fra-01 · Frankfurt512 ms
  ...7 more PoPs
 
▸ deployed to https://hello-world.compute-edge.app
  ttfb at edge: 12-18 ms (depending on PoP)
▸ Primitives

Small surface. Composes into the apps you actually want to build.

Functions
HTTP handlers at the edge

Write a request → response function. We run it within milliseconds of every user. Cold starts under 5 ms.

$ compute-edge deploy ./hello.ts
KV
Sub-ms key/value reads

Distributed key/value backed by per-PoP SSD. Read latency under 1 ms in-region; eventual consistency globally.

› kv.get("session:abc")
Objects
S3-compatible blob storage

Drop in for any S3 SDK. Replicated to every PoP region, served from the nearest one. No egress fees.

› s3.putObject("bucket", "key", buf)
▸ 14 PoPs · 6 regions

Where your code runs today.

adl-01Adelaide, AU
syd-02Sydney, AU
per-01Perth, AU
mel-01Melbourne, AU
bne-01Brisbane, AU
akl-01Auckland, NZ
wlg-01Wellington, NZ
sin-01Singapore
nrt-02Tokyo, JP
hkg-01Hong Kong
fra-01Frankfurt, DE
ams-01Amsterdam, NL
tll-01Tallinn, EE
iad-01Ashburn, US
We replaced three Lambda regions with compute-edge in an afternoon. p99 latency dropped from 340 ms to 38 ms.
Henri Saar · staff engineer, Kalevi Analytics