PRODSovereign European BaaS platformOpen Dashboard →
ACCUEIL/PRODUITS/EDGE FUNCTIONS
EDGE FUNCTIONS · WASM + AI

WASM across 300 POPs,
5ms cold start.

Multi-language WASM runtime (Rust, Go, TypeScript, Python). Crons, DLQs, secrets, streaming, unified AI Gateway. Billed to the exact CPU millisecond.

5ms
COLD START
300
POPS
2.4M
JOBS/DAY
functions/ocr.tsTYPESCRIPT
Deno.serve(async (req) => {
  const form = await req.formData()
  const file = form.get('document')
  // OCR / WASM processing
  return Response.json({ status: 'processed', file: file?.name })
})
#
Feature set

From HTTP handlers to ML-heavy async jobs

WASM multi-runtime

Rust, Go, TypeScript, Python compiled to WASM. Cold-starts under 8ms, memory-safe, fully sandboxed.

COLD START · 5-8ms

Cron scheduler

Standard crontab + extensions (@every 15m, @weekly). Drift guaranteed under 100ms, manual triggers supported.

SCHEDULE · CRONTAB

Job queue + DLQ

NATS JetStream under the hood. Exponential backoff, automated DLQ, one-click replay from Studio.

NATS · JS · DLQ

Streaming responses

SSE, HTTP/3 streaming, response chunks. Ideal for LLM streaming, transcoding, ML inference.

HTTP/3 · CHUNKED

AI Gateway

Unified proxy for OpenAI, Anthropic, Mistral, Llama. Semantic caching, per-tenant budgets, observability baked-in.

UNIFIED · CACHED

Preview deployments (roadmap)

One preview URL per GitHub/GitLab PR, cloned environment variables — planned roadmap feature.

ROADMAP
#
Global network

Deployed across 300 POPs, visible in 12 regions

Your code runs at the POP closest to the user. The 12 core regions host isolated runtimes, while 288 edge POPs route and cache traffic intelligently.

ANYCAST · BGP ROUTING · LATENCY FROM PARIS
par · 5ms
lhr · 8ms
fra · 9ms
iad · 11ms
sfo · 14ms
gru · 18ms
nrt · 22ms
sin · 24ms
syd · 28ms
bom · 26ms
jnb · 30ms
dxb · 24ms
#
Performance

Measured in production, not synthetic benchmarks

5ms
COLD START P50
For a 200 KB Rust WASM handler
300
GLOBAL POPS
Anycast routing, billed per CPU millisecond
99.99%
ENTERPRISE SLA
20× monthly credit guarantee on SLA breaches
#
Comparison

Aurabase Edge · Cloudflare Workers · Vercel

FeatureAurabaseCloudflareVercel
RuntimeMulti-lang WASMV8 isolatesV8 isolates
Cold start P505ms5ms10ms
CronIncluded · 1s resolutionIncludedVia triggers
Job queue / DLQNATS includedNoAdd-on
AI GatewayIncluded · 4 providersNoNo
Preview deploysRoadmapIncludedPro+
BillingCPU ms · GB-RAM-srequests × CPUrequests × compute
#
Use case

2.4M jobs/day · 99.99% SLA

HELVETICA LABS · DOCUMENT PIPELINE · ZURICH

« 2.4 million WASM OCR jobs/day, 99.99% success rate, quarter of the cost vs our previous Kubernetes stack. The AI Gateway saved us 6 months of engineering on semantic caching alone. »

Lucas Steiner
Infrastructure Lead · Helvetica Labs
#
FAQ

Frequently asked questions

Why WASM instead of V8 isolates?+
WASM supports multiple languages (Rust, Go, TypeScript, Python), is memory-safe by design, and strictly sandboxed. V8 isolates enforce JS/TS only, with a broader attack surface (prototype pollution, etc.). Our cold-starts match or beat isolates thanks to AOT compilation.
How does the AI Gateway work?+
A single unified endpoint that proxies OpenAI, Anthropic, Google Gemini. Semantic caching (prompt embeddings), per-project/tenant budget caps, cross-provider timeout fallbacks. Observability included: tokens, cost, latency per invocation.
What about secrets management?+
Encrypted via AES-256-GCM, injected into the execution environment at runtime and never persisted on edge worker nodes.
What is the maximum execution duration?+
HTTP handlers: 60s (Pro plan), 10 min (Enterprise). Jobs: no strict duration limit, retry-safe. Streaming responses: as long as the client stays connected. Idempotency via Idempotency-Key header.
Can we deploy existing code (Deno, Bun, Express)?+
Standard fetch(request) handlers run as-is. Express middlewares can be adapted via a lightweight shim. Full legacy Node applications are recommended for aura-compute (dedicated VMs) rather than edge functions.
What is the exact billing structure?+
Free plan: 100,000 invocations + 100,000 CPU-seconds/month. Pro plan: €0.20 per million invocations + €0.00001 per CPU-ms. Zero charge for idle time (WASM sleeps). AI Gateway billed at pass-through cost + 3% margin.
EDGE IN 1 COMMAND

Deploy your first handler in seconds.

Zero YAML configuration. Zero cold-start latency. Zero vendor lock-in.

No credit card required · 500 MB free · 50,000 MAU