Private beta · now forming the waitlist

Load test in the cloud. Pay only for the traffic you run.

lmn spins up thousands of virtual users in seconds, streams every metric live as the test ramps, and bills by the vu-minuteno seats, no subscriptions, no shelfware. A 30-second smoke test costs cents.

No credit card Early-access pricing Bring your own cloud
You're on the list — we'll email you when your spot opens. Check your inbox to confirm.
0 → 50k
virtual users in seconds, across 12 cloud regions
per-second
billing — scale to zero between runs
you or us
managed cloud, your infra, or wired straight into CI/CD
live
p95, rps & errors stream as the test runs
Real-time

Watch the test as it happens — not 20 minutes after it finished.

Every virtual user streams back to your dashboard the instant it fires. See latency percentiles climb as load ramps, catch a threshold breach the second it happens, and abort a bad run before it burns your budget.

  • Sub-second metric streaming — p50 / p95 / p99, rps, error rate, throughput
  • Live threshold checks that flip red mid-run and can auto-abort
  • Per-region, per-endpoint breakdowns updating in real time
Latency · p95
streaming
p50
88ms
p95
142ms
p99
318ms
Dynamic pricing

Bills like a taxi, not a gym membership.

Forget per-seat tiers and annual commitments. You pay for exactly the virtual-user-minutes you generate, metered per second. Idle weeks cost nothing. A massive launch-day stress test costs what it costs — and you see the running meter before you hit go.

  • Usage-based — priced per vu-minute, metered per second
  • Live cost estimate before every run, hard spend caps if you want them
  • Bring your own cloud credits — run on your infra at cost
This run · estimate
curve · 12m · 3 regions
peak virtual users8,000
vu-minutes consumed~58,000
rate$0.000041 / vu-min
estimated cost$2.38
monthly spend$41 of $200 cap
Smoke test · 50 vu · 30s$0.02
Nightly soak · 500 vu · 1h$1.20
Launch stress · 50k vu · 20m$84.00
Easy to configure

One YAML file. Or paste a request and go.

Describe the load shape, point at your endpoints, set your thresholds — that's the whole config. Drop in a sample request body and lmn templates the dynamic fields for you, so every virtual user sends fresh, realistic data instead of the same payload 8,000 times.

  • Curve, fixed, stages & spike load profiles out of the box
  • Paste-to-template request bodies with per-field fuzzing
  • Version it in git, run it from CI on every PR
checkout.lmn.yaml
# load shape + target, that's itscenario: checkouttarget: https://staging-api.acme.io load:  mode: curve  rate: { from: 100, to: 8000, over: 12m } flow:  - get:  /products  - post: /cart      # body templated ↓  - post: /checkout body:  sku:   {{choice}}  qty:   {{int 1..5}}  email: {{email}} expect:  p95: < 200ms  err: < 0.5%
The moat

Connect your metrics, traces and logs. Let the model find the why.

A load test tells you that p95 spiked. lmn tells you why. Wire in your Prometheus metrics, distributed traces and logs, and our model correlates them against the live run to surface the actual bottleneck — the saturated pool, the slow query, the GC pause — not just a red line on a chart.

Metrics
Prometheus
cpu, memory, pool saturation, queue depth
Traces
OpenTelemetry
span timing across every downstream hop
Logs
Loki · ELK
errors, warnings, slow-query records
Correlation engine
aligns every signal to the run timeline · ranks likely causes
beta
Likely cause: p95 crossed 200ms at 02:48, 6s after the Postgres connection pool hit 100%. Trace a3f9 shows checkout-svc waiting 1.3s to acquire a connection.
correlated across 3 sources · 0.91 confidence · suggested fix: raise pool size or add pgbouncer
Coming soon

Describe the test. The assistant writes it.

Skip the docs entirely. Tell lmn what you want to test in plain English and the assistant drafts the config, picks a sensible load curve, sets thresholds from your historical baselines, and — once it's running — narrates what's happening and what to look at.

  • Natural-language → ready-to-run YAML
  • Baseline-aware threshold suggestions
  • Plain-English run summaries with the correlation findings inline
SK
Stress test the checkout API up to 8k users over 12 minutes, fail if p95 goes over 200ms.
Done — drafted checkout.lmn.yaml with a curve to 8,000 vu over 12m and a p95 < 200ms threshold. Based on last week's runs I'd also watch p99; want me to add it and kick off the run?
Ask lmn to set up a test…
Be first in line

Join the lmn waitlist.

Early access opens in batches. Get in now for founder pricing, a generous free run allowance, and a direct line to the team building it.

No credit card Founder pricing Unsubscribe anytime
You're on the list — we'll be in touch soon.