Portable mixed-precision math, linear-algebra, & retrieval library with 2000+ SIMD kernels for x86, Arm, RISC-V, LoongArch, Power, & WebAssembly

Node.js bindings for controlling nteract runtimed notebooks and Python kernels.

A JavaScript SDK for executing multi-language code in controlled sandboxes, supporting both synchronous and asynchronous modes, as well as multi-language kernels (Python, R, Node.js, Deno/TypeScript, Java/IJAVA, Bash)

Packed GPU struct schemas + Babylon include emitters with optional AssemblyScript kernels

Experimental library for executing Open-CL kernels over Node.JS buffers.

Zero-GC leak diagnostic for the @zakkster/lite-* ecosystem. Owner-tree attribution via lite-signal; pluggable kernels for classifying leaks.

Drop wasmagent sandbox kernels into the Vercel AI SDK as a `tool()` — edge-safe code execution with one capability manifest, no E2B / OS sandbox needed

Drop wasmagent sandbox kernels into the OpenAI Agents JS SDK as a Tool — edge-safe code execution with one capability manifest, no E2B sandbox needed

Drop wasmagent sandbox kernels into the Anthropic Claude Agent SDK — edge-safe code execution with one capability manifest, reusing your MCP server policy

Edit and execute code snippets in the browser using Jupyter kernels

Jupyter Kernels in the Browser

Rayon-like data parallelism for Node.js: a Vite plugin that compiles "use parallel" functions to worker_threads kernels with SharedArrayBuffer + Atomics

High-performance LLM inference kernels for WebGPU

Use in-browser and regular kernels in JupyterLab

Convert PNG images to C arrays (for OS kernels / embedded)

Reusable Convex application kit: foundation kernels, approvals component, zod table boundaries, WorkOS-backed auth function constructors, and audited command protocol.

MCP server for Forge GPU kernel optimization — generate and optimize Triton/CUDA kernels on real H100/A100 GPUs from any AI coding agent

Tiny TypeScript-native tensor library with autograd, compiling to WebGPU. Train small models in the browser without hand-writing kernels, or run a pretrained one frozen for transfer learning.

Real-time neural frame interpolation on raw WebGPU: hand-written WGSL kernels, 2.9 MB model, ~2 ms per generated frame on a mid-range GPU. The runtime behind the Framegen extension.

Alberta Buck monetary and identity kernels as WebAssembly: integer money arithmetic bit-identical to the Solidity contracts, and the BN254 identity/wallet cryptography, for Node and the browser