Your AI team handles the rest — the clarifying questions, the plan, the code. You stay in the conversation, not in the details.
nocodo is an experiment. A small-business app, described in plain English, built by AI agents that coordinate with each other — and run on your machine, not ours. We're not there yet on everything. But close enough to keep building.
Try it now
The interesting part
Most AI coding tools need the largest, most expensive models to function. nocodo takes the opposite bet: constrain the problem so tightly that a model small enough to run on your laptop — under one gigabyte — can write correct, production-quality code.
The stack is fixed: Rust on the backend, TypeScript on the frontend. The patterns are known. Instead of asking an AI to reason from first principles, nocodo shows it a dozen examples of the exact thing you need. The model copies the pattern. Deterministic code does the cleanup.
The result is less creative, and a lot more reliable. We have a 0.8-billion-parameter model writing correct database functions. That's the premise — and it works because of the constraint, not despite it.
How it works
When you open nocodo, you're greeted by a Product Owner — an AI whose only job is to understand what you want to build. It asks about your users, your data, the features that matter, and the ones that can wait. You answer in plain English, or by clicking options when the choices are clear.
When the PO has enough, it hands off to a Project Manager. The PM turns your conversation into a structured plan: a single epic with tasks, each assigned to the right specialist. You never see the mechanics. You just see the plan.
Specialists — backend, frontend, database — pick up their tasks and get to work. Every epic and task has a comments thread, so you can follow along, ask questions, and steer without managing anyone.
Your main conversation partner. Asks about users, data, and features — using text and clickable choices to keep things clear.
The PO hands off to the Project Manager, who produces a structured epic with tasks — assigned to the right specialist automatically.
Backend, frontend, and database agents pick up their tasks. An Engineering Manager reviews technical work before anyone starts coding.
Every task has a comments thread. Stay in the loop, ask questions, and redirect work — without managing the team yourself.
The team
nocodo ships with agents that have specific jobs. The Product Owner handles requirements. The Project Manager creates the plan. Specialists execute. You never coordinate them — they coordinate each other.
Your main conversation partner. Listens, asks clarifying questions, and summarises what you need — in plain language, not technical specs.
Takes the PO's requirements and produces a structured plan — an epic with tasks, each assigned to the right specialist.
Reviews technical tasks before they reach specialists. Makes sure the work is shaped correctly before anyone writes a line of code.
Backend, frontend, database, and UI agents execute their assigned tasks — running locally on small models, not shipped to the cloud.
Honest status
The conversation layer is fully working — PO intake, PM planning, task assignment, comments, and the full agent coordination flow. That part is solid.
The first coding agent — focused on Rust database code — runs reliably on a 0.8-billion-parameter model. Frontend coding agents are next. Auth, deployment, and managed hosting are on the roadmap.
This is experimental software, built in public. It works for the narrow case it's designed for. If you're a developer who finds the approach interesting, or someone who wants to follow along — welcome.