Namzu vs AutoGPT-style runtimes. controllable kernel vs autonomous runner.
The two tools sit at different layers of the agent stack. The short version: Namzu is an agent kernel — runtime, scheduling, sandboxing. AutoGPT-style runtimes is a opinionated autonomous runner. They compose more often than they compete.
Namzu
Namzu is an open-source TypeScript agent kernel and SDK. It is infrastructure: it provides the primitives for building autonomous agents but does not decide what an agent should do, or when to stop. Control stays with the developer.
AutoGPT-style runtimes
AutoGPT and its descendants (BabyAGI, AgentGPT, and similar) are opinionated autonomous runners — given a goal, they generate sub-tasks, pick tools, and iterate until they decide they are done. They package goal-pursuit as a turnkey product, with the agent loop as a built-in feature.
Choose Namzu if…
- →You want infrastructure, not a turnkey autonomous loop.
- →Determinism, observability, and deliberate stop conditions matter more than "set a goal and walk away".
- →You need to embed agent runtime into a larger system, not run it as a standalone product.
- →Process isolation, scheduling, and checkpointing are first-class requirements.
Choose AutoGPT-style runtimes if…
- →You want a standalone product where you describe a goal and the system iterates until it claims completion.
- →You are exploring the autonomous-agent paradigm and a polished demo matters more than runtime control.
- →Your workload is short, single-shot, and tolerant of unpredictable cost and termination.
These are different products with overlapping vocabulary. AutoGPT-style runners answer "build me a goal-pursuing agent"; Namzu answers "give me a runtime to build any kind of agent on". Migration tends to look like: outgrow the autonomous runner, then rebuild on a kernel where you control the loop.
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.