Real-time · p99 latency active
pulse.config.tsTypeScript
1import { pulse } from 'pulse-agent';
2
3pulse.init({
4 cluster: "prod-us-east",
5 sampleRate: 1.0,
6 alerting: true,
7});
p99 Response TimeLIVE
1m5m15m1h
240ms180ms120ms
15:42:0015:47:0015:52:0015:57:00

Three lines. Full observability.

Drop the agent. Watch every server heartbeat, API response, and memory spike across your entire Kubernetes fleet — before your users feel a thing.

Avg Latency
84ms
-12%
Error Rate
0.03%
-0.01%
Active Pods
247
+3
Alerts Firing
2
critical
KD
SR
PL
AM
Trusted by 4,200+ SRE teams
Kubernetes·Prometheus·OpenTelemetry·Grafana·Datadog Migration·AWS EKS·GKE·AKS·eBPF Tracing·Jaeger·Zipkin·gRPC·REST APIs·Kubernetes·Prometheus·OpenTelemetry·Grafana·Datadog Migration·AWS EKS·GKE·AKS·eBPF Tracing·Jaeger·Zipkin·gRPC·REST APIs·
Industry Report · Finding 01

73% of outages are detected after user impact.

By the time your on-call engineer gets paged, the damage is already measured in lost transactions and SLA violations. Pulse fires alerts the moment a metric crosses threshold — not after a 5-minute polling cycle.

Sub-second metric ingestion across all nodes
Multi-condition alert rules with AND/OR logic
PagerDuty, Slack, OpsGenie native integrations
Noise reduction via automatic deduplication
Alert Configuration
2 firing
p99 Latency Threshold200ms
50ms500ms
api-gateway
p99 latency · 340ms
FIRING
2s ago
auth-service
error rate · 2.4%
FIRING
18s ago
postgres-primary
connection pool · 94%
FIRING
1m ago
cache-layer
memory usage · 78%
OK
healthy

↑ Drag the slider to adjust the threshold and watch alerts fire

Industry Report · Finding 02
Trace: a4f2c8d1340ms · slow
8 spans1 error
SPAN
085ms170ms255ms340ms
POST /api/checkout
api-gateway
340ms
validateSession()
auth-service
12ms
getCart(userId)
cart-service
28ms
SELECT * FROM carts
postgres
22ms
processPayment()
payment-svc
280ms
stripe.charge()
stripe-sdk
274ms
updateInventory()
inventory
18ms
sendConfirmation()
email-svc
14ms

MTTR drops 41% with correlated traces.

When your payment service slows, is it the database, the network, or the upstream SDK? Pulse stitches every span into a single causal trace — click any service, see the exact bottleneck in milliseconds.

Root cause identified
stripe-sdk
274ms · 80% of trace
Affected downstream
3 services
payment → email → inventory
Time to identify
8 seconds
vs 22 min industry avg

← Click any span in the trace to expand its attributes

Live Metrics · prod-us-east

Every signal. One pane.

p99 Latency
api-gateway
84ms
↓ 12ms from 1h ago
Error Rate
all services
0.03%
↓ 0.01% · within SLO
Throughput
ingress
12.4k rps
↑ 8% · peak load
Memory Usage
postgres-primary
68%
↑ 4% · watching
CPU Throttle
worker-pool
2.1%
→ stable · 24h avg
Apdex Score
all
0.96
↑ 0.02 · excellent
Zero-config · 2 minute setup

Install the Agent.
Start seeing everything.

One command. Auto-discovers your services, instruments your code, and starts streaming metrics in under 90 seconds.

Terminal
$ curl -sSL https://get.pulse.sh | bash -s -- --cluster prod-us-east
No credit card required
SOC 2 Type II
GDPR compliant
Data never leaves your VPC
99.97%
Uptime SLA
across all regions · 12 months
<1s
Metric Ingestion
from event to dashboard
4,200+
Engineering Teams
from seed to Fortune 500
41%
Faster MTTR
measured across 800+ incidents

From the engineers who run it at 2 a.m.

"

We had a memory leak in our Rust service that took 3 days to find with our old stack. Pulse correlated the trace in 11 seconds. Literally 11 seconds.

MC
Marcus Chen
Staff SRE · DataScale
"

Our enterprise clients need proof of four-nines. Before Pulse, that meant a painful export every quarter. Now I share a live dashboard URL and it's done.

PN
Priya Nair
VP Platform Engineering · CloudOps Pro
"

2 a.m. Kubernetes incident. Three time zones of engineers. Pulse's correlated alerts meant we had root cause before the incident bridge even fully assembled.

JK
Jordan Kowalski
Principal DevOps Engineer · Nexus Systems

Your cluster is already talking.

Pulse listens. Drop the agent and get full observability in under 90 seconds — no YAML archaeology required.