Architecture

How Scalix Cloud is built — gateway, services, and the agent-native platform

Overview

Scalix Cloud replaces 10+ cloud services with one unified platform. One API token authenticates across database, compute, AI inference, storage, functions, and more.

Gateway

Every request enters through the Rust/Axum gateway. It handles:

Authentication — validates API keys (scalix_sk_) and personal access tokens (scalix_pat_)
Routing — proxies to downstream services via service-specific routers
Rate limiting — per-key and per-endpoint limits
Metering — tracks usage for billing
WAF — request inspection and blocking
SSL/TLS — automatic certificate provisioning for custom domains

plaintext

Gateway (:443, public)
├── /v1/compute/*    → Container deployments
├── /v1/ai/*         → AI inference
├── /v1/storage/*    → Object storage
├── /v1/functions/*  → Serverless functions
├── /v1/kv/*         → KV store
├── /v1/events/*     → Event bus
├── /v1/cron/*       → Scheduled jobs
├── /v1/auth/*       → End-user auth
├── /v1/domains/*    → Custom domains
├── /v1/sandboxes/*  → Interactive sandboxes
├── /v1/build/*      → Builds
├── /v1/registry/*   → Container registry
└── /api/v1/*        → ScalixNova database

Each /v1 surface is a capability behind the same token and the same gateway — there is nothing else to wire up, configure, or authenticate against separately.

ScalixNova Database

A serverless PostgreSQL engine where compute is separated from durable storage, enabling instant branching and point-in-time recovery.

Key capabilities: instant branching (under 100ms, copy-on-write), point-in-time recovery, query firewall, and connection pooling.

Compute

Isolated microVMs with fast cold starts. Three execution modes:

Mode	Use Case
One-shot code	Run a script, get output
Interactive sandboxes	File system + shell for AI agents
Container deployments	Long-running services with autoscaling

AI Inference

Scalix AI models behind an OpenAI-compatible API:

Models: the Scalix Lumio family — scalix-lumio-lite (available) through the scalix-candela flagship (coming soon), with automatic retries and high availability
Governance: cost controls, model allowlists, usage tracking
Streaming: SSE-based token streaming for real-time responses

Unified IAM

One token format across all services:

Token Type	Format	Use Case
API Key	`scalix_sk_*`	Server-side SDK usage
Personal Access	`scalix_pat_*`	Developer CLI usage

Fine-grained scopes control access across all services. A compute:deploy scope doesn't grant database:query access.

Cloud Independence

Scalix Cloud deploys on any infrastructure.