v3.2 Now serving 318 metros · 41ms global p50

Your code, already running
in 300+ cities.
Zero config. One command.

Nimbus deploys to the edge in milliseconds, not minutes. Push a build, we handle the rest — routing, replication, rollbacks. No clusters to babysit, no cold starts to hide.

$npm i -g @nimbus/cli
·no credit card
·50GB-egress free / mo
~/projects/payments-api — zsh
connected
Trusted in production by engineering teams at

One deploy. Live everywhere in 4 seconds.

Every push replicates to 318 metros automatically. Requests route to the closest healthy region — if one disappears, traffic heals in under 200ms. No DNS tricks, no manual failover.

318
METROS
41ms
P50 LATENCY
99.99%
UPTIME / 90D
4.2s
BUILD → GLOBAL
Live · 2.1M req/s across network
Primary region Active failover Anycast edge hover a node to inspect

Built for engineers who ship at 3am and sleep at 4.

Everything you'd build yourself if you had six months and a platform team. We did the boring hard part so you can keep the interesting hard part.

Instant rollbacks, no DBA required

Every deploy stores a complete immutable snapshot. One command — nimbus rollback — flips traffic back in 180ms. No schema migrations to reverse, no cache to flush, no 2am Slack thread.

avg rollback time · 180ms · last 30d

Zero config, opinionated correctly

Bring a Dockerfile, a build script, or just a folder of JS. Nimbus infers runtime, health checks, secrets, and routing. Override anything in nimbus.toml — but you won't need to for 90% of services.

convention ratio · 91% zero-config deploys

No cold starts

Warm pools in every metro. First-byte latency under 40ms at p99 — even after 6 hours of silence.

Secrets that travel

Encrypted at rest, scoped per region, rotated without redeploy. Integration with Vault, AWS SM, Doppler.

Observability built in

Traces, metrics, logs — OpenTelemetry-native, no agent install. Query with SQL or click a graph.

Health checks that actually check

Synthetic probes run from 12 geographically distributed sentinels every 5 seconds. If a region degrades — latency drift, error creep, certificate expiry — traffic shifts away automatically and you get a single, deduplicated incident. We don't page you for a flapping node. We page you when users would have noticed.

MTTR · 47s · last quarter · 1,247 incidents auto-resolved

Edge KV & queues

Globally consistent key-value and durable queues. Eventually consistent reads in 12ms, strong reads in 45ms.

From git push to global in 4.2 seconds.

No CI/CD YAML. No Helm charts. No waiting on a build queue while someone's grumpy deploy clogs the runner. Nimbus watches your repo, builds on push, and ships to every region in parallel — rolling out canary-first, promoting on signal.

See the deploy deep-dive →
01
git push origin main
Webhook fires within 80ms. Build context streamed to 3 build regions in parallel.
detected · 412 files · 18.4MB · 1.2s
02
Build & sign
Reproducible build, SBOM attached, image signed with cosign. Cached layers across regions.
build ok · 1.8s · sig: sha256:9f3a…
03
Canary to 3 metros
5% traffic in Frankfurt, Singapore, Ashburn. Health probes sample error rate and p99 latency.
canary clean · err -0.01% · p99 38ms
04
Promote to 318 regions
Parallel rollout with anti-affinity. Old revision stays warm for 6 hours — instant rollback window.
live in 318 metros · total 4.2s

Pick a city. We're probably already there.

Every region runs the full stack — compute, KV, queues, observability. No "edge" tier that's just a CDN with a Lambda taped to it.

Metro
p50
p99
Load
Status
🇩🇪 Frankfurt fra-1
12ms
38ms
62%
healthy
🇸🇬 Singapore sin-1
18ms
44ms
71%
healthy
🇺🇸 Ashburn iad-1
9ms
31ms
58%
healthy
🇯🇵 Tokyo nrt-1
22ms
51ms
83%
elevated
🇧🇷 São Paulo gru-1
28ms
67ms
44%
healthy
🇦🇺 Sydney syd-1
31ms
72ms
55%
healthy
🇮🇳 Mumbai bom-1
24ms
58ms
77%
healthy
🇿🇦 Cape Town cpt-1
36ms
89ms
38%
healthy
View all 318 regions →

Looks like your code. Runs like infrastructure.

No proprietary SDK. No vendor lock-in wrappers. Write a function, export a handler, ship. The runtime speaks HTTP, gRPC, and WebSockets — and you can run the exact same code on your laptop with nimbus dev.

  • Multi-runtime — Node 20, Python 3.12, Go 1.22, Rust, Bun, Deno. Bring your own if it speaks Linux.
  • Local parity — the dev server uses the same runtime image as production. No "works on my edge" bugs.
  • Open standards — OpenTelemetry, OCI images, SPIFFE identities. Export and leave anytime.
handler.ts
nimbus.toml
deploy.sh

Receipts, not marketing.

Measured over the last 30 days across all customer workloads. Numbers updated every minute from our public status page.

41ms
global p50 latency
99.99%
uptime · 90 days
2.1M
requests / second
4.2s
build to global live

"Finally, infrastructure I forget about."

"We replaced a 14-region Kubernetes setup and a full platform team's worth of glue code. On-call pages dropped 80% in the first month. The other 20% were genuinely our fault."
MA
Mira Aoki
Staff Eng · Hyperlane
"The rollback story alone justified the migration. We had a bad deploy at 2am, ran nimbus rollback, and were back live before the on-call finished brewing coffee. 180ms is not a typo."
DK
Dmitri Kovac
SRE Lead · Stackform
"I read the docs once, ran the CLI, and had a service in 47 metros before lunch. It felt like cheating. Then I went looking for the catch. There isn't one — it's just a really well-built platform."
RP
Rae Park
Backend Lead · Pulsewave

Pay for what runs. Not for what you forgot to turn off.

Per-request billing with no idle fees, no egress surcharges between regions, no "talk to sales" for the meter to make sense. See the calculator →

Hacker
$0/mo
For side projects and your first prod deploy. No card required.
  • 50GB egress / month
  • 3 regions of your choice
  • 1M requests / mo
  • Community Slack support
  • No SLA
Start free
Enterprise
Custom
For regulated workloads, dedicated capacity, and on-prem edges.
  • Dedicated regions / BYO cloud
  • SOC 2 Type II, HIPAA, ISO 27001
  • 99.999% SLA with credits
  • Named SRE + Slack channel
  • Private VPC peering
  • Audit logs · 7 year retention
Talk to engineering →

Ship to 300 cities before your next meeting.

Free tier covers most side projects forever. Production starts at $0.40 per million requests. No credit card, no sales call, no "let's hop on a quick sync."

~4 second setup · 318 regions · cancel by closing the tab