MaShop is built around an AI code generator. This charter explains what that AI is, where your prompts go, what its output is worth, and what stays under your control. It works alongside the Terms and the Privacy Policy. Where those documents define a right or a duty, they control. This page explains how the AI actually behaves.
What powers MaShop
When you ask the builder to create or change your store, the reply is produced by large language models. We call the code and content they produce for you the Generated Code.
We do not run these models ourselves. They are open models, meaning models whose weights are publicly released, served for us on third-party inference infrastructure. Our inference provider is Fireworks AI. No other AI provider receives your prompts today. The platform code contains an optional fallback route through a second provider, OpenRouter; it is not enabled, and no prompts flow through it. If we enable it, we will update this charter and the Privacy Policy first. Fireworks appears, with what it processes, in the recipients table of the Privacy Policy.
The specific models change over time as better open models are released, so this charter does not pin itself to model names. The category does not change: open models, served by Fireworks AI.
You are talking to an AI
The builder chat is an AI system, not a human. Every reply in it, and every file it writes, comes from a machine learning model responding to your instructions. The EU AI Act (Article 50(1)) requires that people interacting with an AI system are informed of that fact unless it is obvious. We would rather state it outright than rely on it being obvious.
The AI in the chat can also act. Through its tools it can commit files to your repository, run database migrations on your connected Supabase project, and trigger deployments. It does this in response to what you ask, and the results are visible in your workspace and your repository history.
Where your prompts go
When you send a message in the builder:
- Your prompt leaves your browser and reaches our platform.
- We attach relevant project context: your project details, files the AI has read, conversation history, and stored AI memory.
- We send that package to Fireworks AI, which runs the model and streams back the reply and any Generated Code.
- We store the conversation and the resulting files in your project so you can come back to them.
That is the whole path. Your prompts do not flow to OpenAI or Anthropic. This mirrors section 4 of the Privacy Policy.
What we do not do
We do not sell your prompts or your Generated Code. We do not use your prompts to build advertising profiles or to target ads.
On training and retention, the strongest commitments come from our providers' own terms, not from anything we can technically enforce. Here is what those terms say (retrieved 2 July 2026):
- Fireworks AI's privacy notice states: "We do not use your prompts, training data, or API inputs to train or improve our AI models without your explicit opt-in." Its documentation adds that Fireworks "does not log or store prompt or generation data for any open models, without explicit user opt-in", and that prompt data exists "only in volatile memory for the duration of the request". Every model we use is an open model on Fireworks serverless, so this scope covers our traffic there.
- If we ever enable the OpenRouter fallback, its documentation states: "OpenRouter does not store your prompts or responses, unless you opt in". OpenRouter is a router that forwards requests to third-party model hosts, each with its own data handling policy. We would update this charter before enabling it.
These are the providers' commitments under their own terms. We pass them on so you can weigh them yourself.
Generated Code is a draft
The AI can be wrong. Generated Code can contain bugs, broken flows, security gaps, or assumptions that do not hold in your business. The builder has tools that validate and auto-fix some issues and can scan a deployed site for problems, but no automated loop catches everything.
Before you run a store that takes real money from real customers, review the code and test the full flow yourself, including checkout. Treat every AI output as a draft that you accept, change, or reject. You are the reviewer of record for your own store.
Outputs may also be non-unique. Models produce similar code and text for similar requests, so another merchant's Generated Code may resemble yours. Do not treat generated names, copy, or designs as exclusive. Run your own trademark and licensing checks where that matters.
AI-generated marking
The Generated Code is AI-generated. So are the texts, page structures, and configuration the builder writes into your project. We state that here rather than let anyone assume a human wrote it.
Article 50(2) of the EU AI Act expects providers of generative AI systems to mark synthetic output in a machine-readable format where technically feasible. Technical standards for marking generated source code are still emerging, and we do not claim AI Act compliance on this page. What we state plainly is the fact itself: output from the MaShop builder comes from an AI.
Commerce-aware limits
The builder is commerce-specific. Its tools scaffold a Next.js project, create products and categories, set up coupons, run database migrations, upload images, install packages, deploy the site, attach custom domains, and scan the live site for issues. Some of what it wires touches regulated ground: payment provider configuration, stock and order handling, prices, and legal-adjacent storefront content.
The AI does not know your legal obligations. It cannot judge whether your prices, tax setup, product claims, or customer flows comply with the law where you sell. You remain responsible for your storefront, its content, its compliance, and your agreements with payment providers. The Terms set out this split in detail.
Your ownership and controls
The Generated Code is pushed to a GitHub repository on your own account, created or linked with your authorization. It is yours under the Terms. You can clone it, edit it outside MaShop, and deploy it where you choose.
On our side, we store:
- your builder conversations, per project: the messages you and the AI exchange;
- AI memory entries: short preference notes the AI saves about your project, such as brand rules, saved preferences, and things you told it never to change;
- generation logs and token counts, used for billing and debugging.
Deleting a project also deletes its conversation history. We do not currently offer in-product export or item-by-item deletion of AI memory. Email contact@edeverse.com and we will handle it. Retention periods are in the Privacy Policy.
Updates and contact
We revise this charter when our inference providers, our tool set, or our data handling change in a way that affects what it says. The date shown at the top of this page reflects the latest revision.
Questions, concerns, and reports about the AI features go to contact@edeverse.com. If the AI produced something you believe is harmful or mishandles your data, tell us in that email and describe what happened.