Trace the slow hop
Every p95 spike opens on the trace that caused it, with the offending span already highlighted.
edge · cdn
gateway · auth
orders-svc
postgres · SELECT orders
redis · GET cart
serialize
New · Synthetic checks from 24 regions
Know your API is slow before your customers does.
Relay watches every endpoint from real traffic and synthetic probes, traces the slow hop, and pages the right person with the request that broke.
No card required · 2-minute SDK install · SOC 2 Type II
checkout-apius-east-1 · production
live · 1sp50
139ms
p95
172ms
uptime 30d
99.982%
SLO 250ms
- 200GET /v2/orders124ms
- 201POST /v2/checkout162ms
- 200GET /v2/inventory98ms
Trusted by 2,400 engineering teams shipping 9 billion requests a day
Northwind FreightHalcyonTessellateBrightpath HealthQuarryLumen PayOaklineParcelwise
The product
Every slow request, explained in one screen
Relay connects the metric that moved to the trace that caused it and the deploy that shipped it. No dashboards to build first.
Page the owner, not the channel
Alerts route by CODEOWNERS, so the person who wrote the query gets the page.
p95 above 250ms for 5m
POST /v2/checkout · owner @maya.okafor
MOJTRK
Probes from 24 regions
Synthetic checks every 30 seconds catch the outage your quiet region would hide.
Status pages that write themselves
Incidents post, update and resolve from the same signal that paged you.
45 days ago99.98% uptimeToday
Budgets, not thresholds
Set an SLO once. Relay alerts on burn rate and pins the deploy that spent it.
How it works
From npm install to first insight in under ten minutes
- 01
Drop in the SDK
One package, one line in your server entry. Relay instruments HTTP, gRPC, queues and the database driver you already use.
$ npm i @relay/node import { relay } from "@relay/node" relay.start({ service: "checkout-api" }) - 02
Relay learns your baseline
Within an hour Relay maps every route, learns its normal latency per region and time of day, and proposes SLOs you can accept in one click.
✓ 142 routes discovered ✓ baseline p95 learned for 24 regions → suggested SLO: 99.9% < 250ms - 03
Get paged with the answer
When something drifts, the alert arrives with the trace, the diff and the owner. Most teams resolve in minutes, not war rooms.
[relay] p95 412ms on POST /v2/checkout cause: orders.sql:42 missing index since: deploy #4812 by @maya.okafor
Integrations
Plugs into the tools your on-call already lives in
Forty native integrations, plus OpenTelemetry in and webhooks out for everything else.
- PagerlineOn-call schedules and escalation
- ChatwaveAlerts as threads with one-click ack
- TicketryOpen an issue from any incident
- OpsboardMirror incidents to your war room
- Shipyard CIMark every deploy on every chart
- HookshotSigned webhooks for anything else
- CloudnestAuto-discover services and regions
- VaultkeyRotate ingest keys with your secrets
- GridlogJump from a span to its log lines
- QueuebirdTrace across queues and workers
- DatadeckExport SLOs to your warehouse
- TracewireIngest existing OpenTelemetry
requests watched per day
faster mean time to resolve
probe regions on six continents
median alert-to-owner time
Customers
On-call got quieter. Incidents got shorter.
We used to find out about checkout latency from the support queue. Relay paged us eleven minutes before the first ticket, with the slow query already circled.
The SLO suggestions were closer to right than the ones we argued about for a quarter.
Install took four minutes. Our first real insight took six.
Deploy markers on every chart ended the 'was it us?' thread forever. Now it is always obvious which merge did it.
Routing alerts by code owner cut our pages per engineer in half without missing a single incident.
Pricing
Priced per team, not per host
Start free. Upgrade when Relay has paged you about something that mattered.
Save 20%
Hobby
For side projects and first services.
$0forever
- 3 services
- 1M requests / month
- 5 probe regions
- Email alerts
- 7-day retention
Team
For product teams with an on-call rotation.
$39per seat / month
Billed $468 per seat yearly
- Unlimited services
- 50M requests / month
- All 24 regions
- Owner-based routing
- Public status page
- 30-day retention
Scale
For platforms with strict SLAs.
$159per seat / month
Billed $1908 per seat yearly
- Everything in Team
- 1B requests / month
- SAML SSO and audit log
- Private probe regions
- 13-month retention
- 99.99% ingest SLA
FAQ
Questions from the last on-call retro
Something missing? Our engineers answer in the support chat, usually within the hour.
Under 0.4ms at p99 in our benchmarks. Spans are batched in memory and flushed off the request path every two seconds, and sampling adapts automatically when traffic spikes.
Your next incident is already in the traffic.
Install Relay before lunch and let it tell you where.