Namzu vs Mastra. kernel vs application framework.
The two tools sit at different layers of the agent stack. The short version: Namzu is an agent kernel — runtime, scheduling, sandboxing. Mastra is a typescript application framework. They compose more often than they compete.
Namzu
Namzu is an open-source TypeScript agent kernel and SDK. The surface area is intentionally narrow: process lifecycle, scheduling, memory, IPC, sandboxing, and checkpoint/resume. It does not ship a UI, a deployment runtime, or a workflow DSL.
Mastra
Mastra is a TypeScript-first framework for building AI applications and agents. It bundles agent composition, workflows, RAG primitives, evals, and a developer playground. The scope is broader than a kernel — closer in spirit to a full-stack app framework.
Choose Namzu if…
- →You want a thin runtime layer you can put underneath any framework or your own composition code.
- →You need process-level isolation and checkpoint/resume — capabilities that live below the framework layer.
- →You prefer small, stable interfaces over a large bundled feature set.
Choose Mastra if…
- →You want one framework that covers agents, workflows, RAG, and evals out of the box.
- →You are starting from scratch and value a polished, opinionated developer experience over runtime control.
- →You like a built-in playground/UI and a single mental model for everything.
A Mastra application can run on top of a Namzu kernel. You keep Mastra for composition and ergonomics, and let Namzu handle process boundaries, sandboxing, and checkpointing — concerns Mastra deliberately leaves to the runtime. The two projects share TypeScript-first sensibilities and overlap less in practice than the comparison page implies.
Try the kernel underneath your stack.
Namzu installs from npm and runs on Node.js today. Pair it with whichever agent layer you already use.