navox-agents

Navox Agents

15 AI specialists. One sprint cycle. Zero dependencies. Think. Plan. Build. Review. Test. Ship. Reflect.

One plugin. 15 agents. Full sprint cycle. Install in 30 seconds — no dependencies, no platform, no login.

/plugin marketplace add https://github.com/navox-labs/agents

A Claude Code plugin and a Python CLI for autonomous orchestration. Install the plugin for interactive sprints. Use the SDK to run your team while you sleep.

---

Why I built this

When I heard AI will replace engineers, I wanted to find out how. Not in theory — in practice. How does a non-human replace humans in an engineering capacity?

I'm Nahrin, founder and builder at Navox Labs. Earlier in my life I worked in building architecture — reviewing blueprints, designing structures, making sure everything held together before a single brick was laid. I'm the only architect I know who transitioned to full-time software engineering without a formal CS education. I told myself: once an architect, always an architect. I understood tech through a builder's lens — scalability, optimization, reliable structure.

I've been building physical products for 16 years and tech products for 7. Twenty-three years of building, and I never thought AI could replace my ability to make decisions. It can't. But it can execute them.

Navox Agents is how I ship.

As a solo founder, I needed a reliable team that requires minimum management. Not a chatbot I have to babysit. Not a prompt I cross my fingers on. A team of specialists with defined roles, handoff contracts, quality gates, and the discipline to self-evaluate before passing work downstream.

I believe a single builder with the right tooling can move faster than a traditional team. This is my answer.

Fully open source. A Claude Code plugin for interactive sprints and a Python CLI for autonomous orchestration. I combined these two because I want this to reach as many people as possible. These agents are built with best practices from Anthropic themselves. They self-evaluate for high reliability. They're orchestrated and work exactly like a team of specialists.

This is an open-source software factory. I use it every day. I'm sharing it because these tools should be available to everyone.

Fork it. Improve it. Make it yours.

And if you want to hate on free open-source software — you're welcome to, but I'd rather you just try it first.

---

Who is this for

Founders and CEOs — especially technical ones who still want to ship. Run a full sprint while you focus on strategy, fundraising, or sleep.

First-time Claude Code users — structured roles instead of a blank prompt. Instead of figuring out what to ask Claude, just tell the team what to build.

Tech leads and staff engineers — orchestrated specialist agents that follow your architecture decisions, not fight them. Sprint chains enforce the same discipline you'd expect from a senior team.

Solo builders — one person, 15 specialists. The math works.

---

What happens when you run it

$ /agency-run FULL Build a cookie clicker game with crab theme

→ Strategist: "Single HTML file is smart — no build step, instant deploy.
   But cookie clickers need prestige mechanics to retain players. Adding to spec."
→ Spec Writer: 7-section spec with 12 acceptance criteria
→ Architect: Component tree, state machine, CSS custom properties for theming
→ Gate 1: You review. You approve.
→ UX + Security: Running in parallel...
→ Full Stack: 1,330 lines written. All acceptance criteria met.
→ Checkpoint: App running at localhost:3000 — LGTM / FEEDBACK / STOP?
→ Reviewer: 7 specialists. 2 issues found. Auto-fixed.
→ QA: 6 bugs caught. 6 bugs fixed. 14 tests passing.
→ Security: No vulnerabilities. Auth N/A. Cleared.
→ Shipper: v1.0.0 tagged. PR #1 created.
→ Retro: 4 learnings captured to project memory.

Total time: 7 minutes. Human edits: 0.

---

See it work

nom.sh — one prompt, 7 minutes:

A crab cookie clicker. 1,330 lines. 6 bugs caught by QA. Zero human edits. See the code