TypeGPU
TypeGPU is a modular, type-safe WebGPU toolkit that lets developers write GPU kernels and shaders in TypeScript, auto-generate WGSL, and manage strongly typed buffers. It reduces boilerplate while preserving low-level control and interoperates cleanly with existing WebGPU code and libraries.
Overview
Define a TypeScript function with ‘use gpu’, declare typed buffers, and compose kernels as reusable modules. Resolve functions to WGSL when you need the shader text, or build guarded compute pipelines and dispatch threads directly—while retaining the option to unwrap resources for vanilla WebGPU.
Typed shader authoring without losing WebGPU control
TypeGPU suits graphics engineers, web developers exploring GPU compute, visualization authors, simulation and generative-art practitioners, and library maintainers who need reliable interop. Teams building engines, editors, or research prototypes can enforce buffer schemas, pass typed data between modules, and avoid costly CPU round-trips. It’s particularly useful when different libraries must share GPU resources safely without abandoning types or duplicating memory.
- Author TypeScript functions that compile to WGSL with strong, inferred types.
- Define schema-typed GPU buffers and dispatch compute pipelines without boilerplate.
- Eject to raw WebGPU resources whenever you need fine-grained control.
- Compose reusable shader modules and generate dynamic code safely at runtime.
- Integrate with existing libraries while preserving types and zero-copy data flow.
Why engineers choose TypeGPU
Who should use TypeGPU
Install the core package with your package manager and initialize a root in your app. Write shader functions in TypeScript and mark GPU sections with ‘use gpu’. Define buffer schemas, allocate typed resources, and either resolve WGSL for inspection or build guarded compute pipelines and dispatch. Use the bundler plugin to enable authoring directly in JS/TS, and the CLI to generate types and layouts automatically. Explore the official documentation, examples, and benchmarks to understand patterns, performance considerations, and integration strategies with existing WebGPU code.
Write once in TypeScript, run on the GPU with confidence—without surrendering raw WebGPU control.
Getting started
TypeGPU combines a practical type system with a low-level, WGSL-aligned model, making it easy to scale from prototypes to production without lock‑in. Typed schemas, safe module composition, and optional ejection points keep teams productive while preserving full WebGPU control, supported by plugins, codegen, and focused utility packages.
Open the tool and review its core product experience.
Create your account or access your existing workspace.
Use your own task to judge speed, quality, and fit.
Check similar AI tools before making a final decision.
TypeGPU
Software Mansion is a software agency and open-source contributor known for deep expertise in performance-critical web and mobile technologies. As Core React Native Contributors, they build tools and libraries that emphasize developer ergonomics, reliability, and low-level control for demanding applications.
TypeGPU is a modular, open-ended toolkit for WebGPU that brings advanced typing to GPU programming in the browser. Developers author shader logic as TypeScript functions and mark GPU regions with a ‘use gpu’ directive; TypeGPU then generates equivalent WGSL and manages typed resource layouts. Its low-level design mirrors WGSL constructs, helping you learn and apply WebGPU fundamentals without losing productivity. Strong type inference propagates across kernels, buffers, and pipelines, catching shape and type mismatches early while enabling safe composition. You can always unwrap typed handles to raw WebGPU resources for fine-grained control and straightforward interop. The ecosystem includes utility packages for colors and noise, a bundler plugin to enable shader authoring in JS/TS, and a CLI for automatic code generation. Use it as a foundation for new apps or adopt pieces incrementally within existing engines and libraries.
Visit TypeGPU
Comments (0)
No Comments Found