latch

api gatewaycli-nativeedge deploys↳ latch v2.4.0

( core primitives )

Single binary

One 18MB binary — install with curl, no runtime, no dependency tree to audit.

Edge routing

Requests route through 34 edge regions with sub-12ms added latency, p99.

Structured logs

Every request emits structured JSON — pipe straight into your existing stack.

Zero-downtime deploys

Atomic rollouts with automatic rollback the moment error rates spike.

( the workflow )

01

Define routes as code

Write your API surface in a single latch.yaml — routes, auth, rate limits, and retries all version-controlled next to your service.

api.latch.dev/routes
02

Deploy from the CLI

`latch deploy` builds, ships, and health-checks a new revision — usually inside 4 seconds, no dashboard required.

api.latch.dev/deploy
03

Trace every request

Full distributed traces down to the database query, searchable by status code, latency, or customer ID.

api.latch.dev/trace
04

Alert before customers notice

Anomaly detection on latency and error budgets pages you in Slack or PagerDuty before SLAs break.

api.latch.dev/alerts
api.latch.dev/routes

( status )

Ninety days, tracked to the millisecond

api.latch.dev99.98%
operationaldegradedoutage

( live dashboard )

Every request, one glance away

requests per second

▲ +21.4%

cache hit rate

94%

edge cache, last 24h

active api keys

▲ +4.7%

p99 latency (ms)

▼ -12ms

sla uptime

99%

trailing 30 days

error rate

▼ -0.01%

v2.4.0Jun 24, 2026

Edge tracing goes GA

  • NewDistributed tracing across all 34 edge regions.
  • New`latch logs --follow` streams production output straight to your terminal.
  • ImprovedCold starts cut 40% on Node and Python runtimes.
v2.3.2Jun 10, 2026

Reliability pass

  • FixResolved a rare double-delivery on webhook retries.
  • Fix`latch rollback` now restores environment variables correctly.
  • ImprovedDashboard search now indexes in under 200ms.
v2.3.0May 28, 2026

Rate limiting rewrite

  • NewPer-route rate limits configurable directly in latch.yaml.
  • ImprovedRouting table now propagates globally in under 900ms.
  • FixCorrected handling of chunked transfer encoding.

( pricing )

Hobby

$0/ mo

For side projects and prototypes — enough headroom to actually ship something.

  • 100k requests / mo
  • 1 edge region
  • 7-day log retention
  • Community support
Start free
Popular

Team

$49/ mo

The default for production APIs — usage-based beyond the included quota.

  • 5M requests / mo included
  • 34 edge regions
  • 90-day log retention
  • Distributed tracing
  • Slack + email support
Start on Team

Enterprise

Custom

Dedicated infrastructure and SLAs for teams running APIs at serious scale.

  • Unlimited requests
  • Private edge regions
  • SSO & audit logs
  • 99.99% uptime SLA
  • Dedicated Slack channel
Talk to sales

( faq )

No — Latch sits in front of your existing service as a reverse proxy. Point your DNS at us, add a latch.yaml, and routing, auth, and observability work immediately.

( get started )

Install the CLI, run `latch init`, and push your first route to production before your coffee finishes brewing.

Install the CLIcurl -fsSL latch.dev/install | sh — free tier, no credit card
api.latch.dev/routes/new