Docs
Everything you need to take a project from an idea to production with super_artifex.
Introduction
super_artifex is a team of AI agents that works like an engineering firm. It researches your idea, writes the product requirements, designs the system, builds the screens and the backend, deploys the app and keeps it running.
The work moves through nine stages. Each stage ends with something you review, and the next stage starts only after you approve it.
Your first project
- Top up your wallet. Research is the first stage and the cheapest.
- Create a project and describe the idea in your own words. Add links, documents or screenshots if you have them.
- Answer the agent's clarifying questions. It shows an estimate before any work starts.
- Read the research brief, then review the PRD that follows from it.
Reviews and approvals
Every review opens in a workspace with the output on the left and a chat with the agent on the right. Approve a section, or select Discuss to ask about it or correct it.
Before changing anything, the agent restates what it understood and gives an estimate. Turn on Explain to have it walk you through the whole output, and Voice to hear it and reply by speaking.
Wallet
Work is paid from a prepaid wallet in naira. Top up with Paystack by card, bank transfer or USSD. Every charge is linked to the project, stage and task that used it.
When the balance runs low you get an alert. If it runs out, work pauses at the next safe point and continues after you top up.
Agent slots
A slot is one agent working at a time. Your account includes three. When all slots are busy, new tasks wait in a queue. You can buy more slots from your wallet.
Deployments
Connect your own server or a DigitalOcean account. The agents set up the server, CI/CD, HTTPS and backups, then release each approved version and check that it is healthy. You can roll back any release.
Bug-report links
Each live project has a bug-report link you can share with testers and users. They do not need an account. A report goes straight to the agents, who try to reproduce it and fix it.
The reporter gets a reference like TKT-0142 and a status page. If the agent needs more detail, it asks the reporter on that page.