Helix is the runtime, telemetry and deploy primitive backend teams use to ship resilient systems — without operating one. Twelve regions, one binary, zero ceremony.
Each panel is a real subsystem — the choreography below mirrors what the runtime does in production. No demo magic.
Helix is opinionated about the primitives, not the application. The same runtime serves a marketing API and a 9-figure trading pipeline.
Drop Helix in front of any origin — REST, gRPC, GraphQL — and pick up rate limits, signed routes and per-region failover without touching your services.
Schedulable, idempotent jobs with first-class retries, dead-letter inspection and checkpointing — the parts every team rebuilds badly.
Code that runs in milliseconds at the edge with cold-start budgets you can actually measure. No mystery, no hidden cron, just primitives.
Instrumentation is automatic; the cost model isn't punitive. Drill from a slow request into a downstream span without opening four tabs.
Your services run on the data plane closest to each request. The control plane is invisible — schedule, rotate keys, pin versions from the CLI or the dashboard.
We benchmarked the same Node service on Helix and on a typical Lambda + ALB + CloudWatch stack. Here is what changed.
“We deleted a quarter's worth of platform tickets the week we cut over. The runtime feels designed by people who've actually been on call.”
No per-invocation. No egress between Helix services. No idle charge for warm runtimes. The estimator on the docs is the model, not a teaser.
Engineering teams ask sharp questions. Here are the ones we hear most often, answered without the marketing layer.
Helix is a programmable runtime, not a function host. You compose primitives — routing, queues, KV, telemetry — instead of stitching four services. Cold start, retention and failover are first-class, not add-on bills.
Yes. Helix accepts WASM modules, container images and a managed Bun/Node runtime. Anything that responds on a port can be promoted to an edge service in one CLI invocation.
You pay for CPU-time used while requests are in-flight. There is no per-invocation fee, no egress between Helix services, and no idle/overhead surcharge. The estimator on the docs site is the model — not a teaser.
Scale-tier customers can run a private Helix control plane in their own VPC or bare-metal racks. The data plane is the same binary that powers the public edge.
SOC 2 Type II and ISO 27001 are in-scope today. HIPAA BAA is available on Scale. Per-region residency is enforced at the routing layer; data never leaves the regions you allow.