What Entirius Is

In Short

Open-source modular ecommerce platform: 40+ Django backend modules on PyPI, a Vue 3 CMS and a React storefront (Next.js). Pick what you need, skip the rest. The platform is open source; the business around it is running store operations for clients, paid from the result. Every module ships with an AGENTS.md that works with any AI coding assistant. Yes, we let robots write our code. No, we’re not sorry.

Why We Built It

The platform grew out of ecommerce projects for clients, from minimal B2B catalogs with no prices and no stock, to full-blown global platforms with multi-currency checkout, regionalized content, and complex fulfillment chains.

No existing platform covered that range well. Something was always missing, or a large part of the codebase had to be ripped out before real work could start, which cost time and money. And even after all that surgery, the result was still too monolithic. Change one thing, break three others.

So we built our own.

There is a second reason, and it is the more important one: we run store operations for clients, and client stores run on this platform in production. We wanted a stack with full control: no vendor lock-in, no licensing games, no waiting for someone else’s roadmap.

That is also why the platform is open source. The business around it is running store operations for clients; AI agents do the repetitive work and our experts approve every change. The apps used in that service are not open source and not published. The platform runs without them.

What It Is Made Of

A modular ecommerce platform. The backend (Volkanos) is a collection of independent Django packages (PIM, checkout, accounts, email, delivery points, and more), each versioned and released from its own repository to PyPI. The frontends (Cynthia) are a Vue 3 CMS for content management and a Next.js (React) storefront for customers. The whole platform runs locally from entirius-zeno, the Docker harness. The product names are explained on Concepts.

Volkanos runs client stores in production, with real traffic and real transactions.

Pick the modules you need. Skip the rest. Each module has its own models, schemas, services, and API. Modules install separately, and each one declares the modules it needs: Matrix, the storefront read model, requires PIM.

The Problem

Ecommerce keeps expanding. The surface area grows every year, and the pool of people who truly understand it end-to-end stays small. The scarce resource is that expertise, not frameworks.

Entirius gives you a platform where each module encapsulates domain knowledge. The agentic workflow means every module comes with structured documentation that both humans and AI tools can use to extend, customize, or debug.

How We Build It

Every Volkanos module goes through an agentic development workflow: specialized AI agents handle architecture review, implementation, testing, and documentation, each within its own domain.

The result for each module:

• Layered architecture (models → services → API) with strict dependency direction
• Pydantic schemas for all request/response validation
• OpenAPI docs generated from the code, and an ER diagram in the modules that keep a docs/erd.md
• Tests at every layer: unit, API, BDD
• AGENTS.md files describing architecture and conventions

Every module ships with an AGENTS.md written for any AI coding assistant. What is published and what is not: Status.

AI-First, Human-Led

The platform is developed by AI agents working inside human-designed boundaries: module contracts, AGENTS.md context, specs generated from code, an end-to-end test harness. Humans own the architecture, review every merge, and decide what ships. The agents just type faster. And the discipline pays twice: a codebase an agent can navigate is a codebase you can navigate, and the same clarity serves both.

The same model is where the platform is heading. An ecommerce platform by itself is a foundation: a solid, boring foundation that every store needs and no store wins on. What a store wins on is operations: catalog quality, pricing, content, fulfillment decisions, the daily work that today consumes human attention. Our goal is a store where agents do most of that operational work and a person approves it, through the same public API every integrator uses. If an agent can run a store through the API, so can you and so can your tools.

That’s the direction, not a product announcement. When agent-driven applications are ready, they will appear in the entirius-app-* family.

And that is also the honest answer to why open source: we give away the foundation and build the operations service on top of it. A commodity base gets better when more people run it, break it and fix it. The value we keep is knowing how to run a store with agents and people.

Platform Overview

Start with what you need. Add modules as the business grows. Muted items are planned, not yet available.

Open Source

The platform is open source under MPL-2.0. The module and SDK repositories are public in the entirius org. Which parts are stable and which are early access: Status. Contribution rules live in the Handbook.

Entirius Apps

The planned application layer on top of the platform: products built around AI agents, shipped as entirius-app-* repositories and used in the operations service. They automate the work that today needs manual expertise (product descriptions, SEO, pricing recommendations, content) and every change goes through a person who approves it. Not embedded agents: separate applications that talk to the platform through the same REST API as everything else. Not open source and not published.

Community

• CONTRIBUTING.md: how contributions are reviewed, and what the maintainers promise and do not promise.
• SECURITY.md: report a vulnerability privately, never in a public issue. The channels are on the Security page.
• SUPPORT.md: where to ask a question, report a bug or propose an idea.
• CODE_OF_CONDUCT.md: the conduct expected in every repository of the organisation.

Next

System Architecture