nexus.os
Back to Blog
Nexus OS hero background — circuit board pattern with blue glow
Announcementnexus-oslaunchopen-source

Introducing Nexus OS: The Orchestration Layer for AI Agents

Leonidas Esquire WilliamsonApril 12, 20264 min read

Why We Built Nexus OS

The AI agent ecosystem is exploding. Every week brings a new framework, a new wrapper, a new way to chain LLM calls together. But production agent systems need more than prompt chains — they need orchestration.

Nexus OS is our answer. It's a single binary, written in Rust, that gives you everything you need to run AI agents in production:

  • Erlang-style supervisors that restart crashed agents automatically
  • WASM-sandboxed skill modules for deterministic, near-zero-cost execution
  • Token-cost optimization that routes tasks to the cheapest capable handler
  • A built-in dashboard at localhost:4200 for real-time monitoring

The Three-Layer Architecture

Nexus OS organizes agent systems into three distinct layers:

  1. Execution Layer — WASM sandboxes, containers, and edge runtimes
  2. Orchestration Layer — Supervisors, sagas, workflows, and pools
  3. Intelligence Layer — Broker routing, cost control, and model selection

Each layer is independent but composable. You can use just the supervisor system, or the full stack.

Getting Started

cargo install naos
naos init my-project
naos create researcher --template research
naos run researcher

That's it. Five minutes from zero to a running agent with automatic supervision, cost tracking, and a web dashboard.

What's Next

This is Phase 1. We're shipping the core CLI, supervisor strategies, the broker routing engine, and the dashboard. Phase 2 will bring the WASM skill marketplace, edge deployment to Cloudflare Workers, and CRDT-based state management.

We're building Nexus OS in the open. Star us on GitHub, try the CLI, and tell us what you think.

Back to Blog