bext/The engineBEXT / 01
The engine

Under the hood.
Ahead of the stack.

A compiler, a V8 execution pool, and a delivery layer. Explore the Rust engine that takes your application from source to response.

Read the architecture
FIG. 01 / ENGINE
Another side of the engine.
IN THIS CHAPTER01Rust core02V8 isolates03Streaming SSR
What's inside

Six layers, one binary

Every layer is written from scratch in Rust, built for speed and isolation — with no native dependencies to wire together.

Rust core

An async, GC-free, production server: routing, automatic TLS, compression, ISR. One binary, zero runtime dependencies.

tsc-rs compiler

Our own TypeScript/JSX compiler in Rust — type-stripping + JSX transform. WASM-portable: it runs in the browser too.

V8 isolate pool

Your loaders and components run in shared, supervised V8 isolates — warm, pooled, with a circuit breaker and render budget.

route_css

A Rust Tailwind generator (encre-css) emits exactly the utilities each page uses, at compile time. WASM-portable as well.

Streaming SSR + ISR

Native TSX through PRISM, with no server-side React: TTFB streaming, Suspense, incremental static regeneration, and cache-key normalization.

Sandbox & security

WAF, rate limiting, per-site isolation, sandboxed execution. Every workload is contained.

Current measurements

Fast where it counts

15.7µs
compiled PRISM · average · 38.3 KB fixture
<1ms
incremental transpile
8isolates
shared V8 pool
2.8MB
compiler in WASM
Portable

The same engine runs anywhere — even in a tab

Because the compiler and CSS engine are pure Rust, we compile them to WebAssembly. play.bext.dev is the proof: it compiles and renders real bext code entirely in the browser, using our actual toolchain. Server, edge, CI, browser, agent sandbox — the same engine, everywhere.

See the engine run in your browser →
Embeddable

Built to drop into your product

The engine is a clean, ownable dependency — not a service to call. Ideal for platforms that need to execute and render generated code, safely and at scale.

Code-gen sandboxes

Compile and run AI-generated code in disposable isolates, with renders measured in milliseconds.

App preview

Give users an instant, live preview of the app an AI just wrote — like our playground.

Agent environments

A contained, fast runtime for agents to build, test, and serve real applications.

Edge & self-host

A single binary to deploy anywhere, or a library to embed — with no external runtime.

Build on it. Or embed it.

Free to run, self-host, and distribute. An OEM license to embed it in your product or AI platform.