NocoBase: When AI Agents and Humans Build Software Together

You've probably tried a no-code tool, hit a wall, and gone back to writing code by hand. Or maybe you've watched an AI coding agent generate a pile of code that works but that you can't easily maintain or adjust. What if the two approaches actually worked together instead of competing? NocoBase is an open-source platform that tries to answer that question by putting AI agents and a visual builder on the same infrastructure.

What It Does

NocoBase is an open-source AI plus no-code platform for building business systems quickly. The core idea is that AI doesn't generate everything from scratch. Instead, it works on top of production-proven infrastructure and a WYSIWYG no-code interface, so you get speed without throwing away reliability.

The platform gives you two ways to build. On one side, coding agents get a full CLI and skills, and they can handle setup, development, migration, and release end to end. It works with mainstream agents like Claude Code, Cursor, Codex, OpenCode, and TRAE. On the other side, people get a WYSIWYG interface where they can visually review and configure data models, pages, workflows, and permissions. You can switch between usage mode and configuration mode with a single click.

The interesting part is that these two modes aren't separate products bolted together. They operate on the same system, which means you can split the work however you want. AI can create data models, pages, and workflows fast, and then a person can refine the UI and interactions visually. Or a person can set up the structure and let AI continue from there.

Why It's Cool

• The collaboration model is the real feature. Plenty of tools claim to be "AI-powered" by wrapping a chatbot around a form builder. NocoBase takes a different angle: it treats AI agents and human builders as two participants working on the same project. The README explicitly frames this as "AI and people build together," and the practical result is that you don't have to choose between the speed of an agent and the control of a visual editor.

• It meets you where you already work. The CLI install is a single npm command, and the agent integration covers tools you're probably already using (Claude Code, Cursor, Codex, OpenCode, TRAE). You don't have to learn a proprietary AI workflow. You run the agent you like, and it talks to NocoBase.

• Built for non-developers too. The WYSIWYG interface is designed for regular users, not just developers. That's a meaningful distinction. If a business user can review and adjust a data model or a permission rule visually, you cut out a whole category of back-and-forth that usually lands on an engineer's desk.

• Business context for AI employees. The README mentions that NocoBase includes AI employees so AI can work directly inside the system, and that those employees get business context automatically. That's a different scope than "help me build the app." It's AI operating inside the app you built, which is a more ambitious (and more useful) target for a business platform.

How to Try It

Getting started is a couple of terminal commands. You'll need Node and npm available.

# Install NocoBase CLI
npm install -g @nocobase/cli
nb --version

# Install a NocoBase app
nb init --ui

# Optional: build together with an AI Agent
codex # claude, opencode

From there, the README points to two detailed guides: one for installing a NocoBase app and one for AI agent integration. If you'd rather poke around before installing anything, there's an online demo at https://demo.nocobase.com/new, and the docs live at https://docs.nocobase.com.

The repository is at https://github.com/nocobase/nocobase. Release notes get posted weekly on their blog if you want to track what's changing.

Final Thoughts

NocoBase is aimed at teams building internal business systems — the kind of data models, pages, workflows, and permission setups that tend to sprawl across a dozen tools. If you're a developer who's been skeptical of no-code platforms because they trap you in a visual editor with no escape hatch, the CLI and agent integration are worth a look. If you're on the business side and want to build without waiting on engineering, the WYSIWYG side is there for you. The honest caveat is that a platform this broad takes some setup to evaluate properly, so start with the demo before committing. The direction here — AI and humans sharing the same build surface — feels like where a lot of tooling is heading, and it's useful to see an open-source project taking it seriously.

Follow @githubprojects for more developer tools and open source projects.