Open Source Edge Runtime

COLD
STARTS
ARE DEAD.

KRNL is a Rust-based edge runtime that deploys serverless functions with sub-5ms cold starts. 200+ edge locations. Zero vendor lock-in. Apache-2.0 licensed.

<5ms
Cold Start P99
214
Edge Locations
47K
Github Stars
2.1M
Deploys / Week
$ krnl deploy --edge --region=auto
Compiling function... done (340ms)
Bundling assets... 12.4 KB
Deploying to 214 edge locations...
 
✓ Live at fn.krnl.dev/api/users
✓ Cold start: 3.2ms (p99)
✓ Memory: 1.8 MB peak
 
$ krnl bench --rps=10000
Sustained 10,000 rps at p99 <8ms
Zero dropped requests
✓ All checks passed
Vercel Cloudflare Supabase Linear Resend Cal.com Trigger.dev Neon
// Capabilities
V8 Isolates
Each function runs in a lightweight V8 isolate. No container boot. No JVM warm-up. Just your code, running in microseconds.
Runtime
Global Edge Mesh
214 edge locations across 6 continents. Requests route to the nearest node automatically. Failover in <50ms.
Network
KV + Durable Objects
Built-in key-value store with strong consistency. Durable Objects for stateful workloads at the edge. No external database needed for simple state.
Storage
WASM + JS + TS
First-class support for JavaScript, TypeScript, and WebAssembly. Bring Rust, Go, or C++ compiled to WASM. No transpilation overhead.
Languages
Zero Config Deploys
One command. No YAML. No Terraform. Push to main and your function is live at the edge in under 4 seconds. Git-native by default.
DX
Apache-2.0
Fully open source. Fork it, self-host it, audit it. No proprietary runtime lock-in. Your functions run on standard Web APIs.
License
// Benchmarks
Measured 2026-Q1 · wrk2 · us-east-1 · 1KB JSON response
Runtime Cold Start (p99) Warm Latency (p50) Max RPS Memory
KRNL FASTEST 3.2ms
 0.8ms
 48,200
 1.8 MB
Cloudflare Workers 4.8ms
 1.1ms
 42,100
 2.1 MB
Deno Deploy 12ms
 2.4ms
 31,800
 4.2 MB
AWS Lambda 127ms
 5.6ms
 18,400
 52 MB
Vercel Edge 8.4ms
 1.9ms
 35,600
 3.1 MB
Ship to the edge in 4 seconds. curl -fsSL https://krnl.dev/install | sh