Engineering role
AI Rapid Prototyper
An AI MVP builder that turns an idea into a clickable prototype on a standard stack in days — and writes down what the prototype is meant to prove.
Describe the idea and the one thing you need to learn, and it builds the smallest version that can answer it: the core flow and nothing else, on a stack chosen for speed — Next.js, Supabase or Neon for data and auth, a component library instead of custom design. It writes the hypothesis and the success measure before the code, adds analytics from the first screen, and keeps a list of what was faked so nobody mistakes the prototype for the product.
Things you could ask for
Written the way you would actually say them. The agent plans the steps itself.
- “I want to test whether freelancers will pay for automatic invoice chasing. Write the hypothesis, the one flow to build and what number would tell us yes or no.”
- “Build a Next.js prototype with Supabase sign-in: a form to add a client, a list of open invoices and one “send reminder” button. Use shadcn/ui, no custom styling.”
- “Add PostHog events to every step of the signup flow in this prototype so we can see where people stop.”
- “Turn this Figma-style description of three screens into working pages with fake data, so we can put it in front of five users this week.”
- “Read the prototype repository and write the list of shortcuts we took — hard-coded values, missing validation, no tests — before anyone calls it production.”
What you get back
A one-page brief with the hypothesis, the flow and the success measure; the prototype’s code as files or a repository; a database schema applied to a Supabase or Neon project you own; analytics events wired to PostHog; and a shortcut list that becomes the issues to fix if the idea survives.
Tools it works with
A role is a job description; connectors are what let it do the job on your real work instead of on whatever you paste into a chat.
Where it is the wrong tool
- A remote agent writes the code but cannot run it. There is no Node, browser or terminal behind it, so it cannot start the dev server or click through the result; a local agent with shell access in the desktop app can install, run and commit on your own machine.
- It does not host the prototype. There is no Vercel or Netlify connector; you deploy it, or run it locally for user sessions.
- A prototype is built to be thrown away. It skips tests, error handling and security review on purpose, so do not put real customer data or payments through it.
- It cannot tell you whether the idea is good. It can make the test fast and honest; the answer comes from the people who use it.
Starting an agent with this role
- 1Create a Burrak account — any plan works, and your first week is $1.
- 2In the Marketplace, open the Roles tab, find Rapid Prototyper and choose “Start an agent as Rapid Prototyper”. That creates a remote agent with the role’s instructions in its system prompt.
- 3Connect the tools it needs — GitHub, Supabase, Neon, PostHog, Linear, Notion — from Connectors.
- 4Give it a brief. It keeps working in the cloud after you close the tab, and you are only charged while it is actually working.
The Rapid Prototyper role, in practice
- How fast is “rapid”?
- For a single flow on a standard stack, the code is usually a day or two of back-and-forth, not weeks. What takes the time is deciding what not to build, which is why it writes the hypothesis before it writes anything else.
- Can the prototype become the real product?
- Parts of it can. The stack is chosen so the schema and the core flow can carry on, and the shortcut list says exactly what has to be rebuilt. Hand that list to the Backend Architect or the Frontend Developer when the idea earns it.
- Do I need to know how to code?
- To run and deploy it, a little — or a desktop app with a local agent that can run the commands for you. To decide what to build and judge the result, no; that part is a conversation.
Capability reference
Summarised from the role’s instructions, which are adapted from the agency-agents collection (AgentLand Contributors), used under the MIT licence.
- Working prototypes and MVPs of a core flow in days, not weeks
- Speed-first stacks: Next.js, Supabase, Prisma, component libraries
- Hypotheses and success criteria written before the build
- Analytics and feedback collection from the first version
- Modular code that can evolve, with the shortcuts documented
Your AI assistant is waiting.
Put it to work today.
Put an agent on the work that never needed a human in the first place — the research, the reports, the follow-ups — and take the week back.
$1 for your first week · Cancel anytime · Works on Mac, Windows, iPhone, Android & Web