The First Product Agent That Ships Whole Apps
Remy is a product agent: describe an app in plain language and it builds the full stack—backend, database, auth, frontend, deployment. Here's how it works and what it costs.
What is Remy?
Remy is a product agent: an AI that builds, ships, and runs a complete application, not just snippets of code. You describe the app you want in plain language. Remy writes a plan for it, compiles that plan into a full-stack app (backend, database, auth, frontend, and deployment), tests it in a real browser, and hands you a live URL.
That’s what sets it apart from the AI tools you already know. A coding assistant edits code in a project you already have. A prompt-to-UI builder makes you a screen. Remy takes an idea and returns a working product. The plan is the source of truth, so you change the app by editing the plan and recompiling, not by hand-maintaining code. (For the category itself, see what is a product agent.)
TL;DR
- Remy is a product agent: you describe an app in plain language and it builds the whole thing: a backend, database, user logins, a website, and a live link you can share.
- The plan is the source of truth: Remy drafts a spec from your description, you approve and tweak it in plain language, and the working code is generated from it.
- It builds the full stack in one step, not just a frontend, so the app can actually save data, log users in, and take payments.
- It works like a team of specialist sub-agents (design, backend code, architecture review, browser testing), not a single chatbot.
- It’s open source and runs on a hardened production platform, so every app inherits 200+ models, 1,000+ integrations, databases, auth, and deployment with no setup.
- Pricing is simple: $99/month ($79 with annual billing) after a 7-day free trial, plus pass-through inference at provider rates with no markup, about $100 to build a typical app.
- Use Remy to build a new app from an idea; use a coding agent like Cursor to edit an existing codebase.
How does Remy work?
Remy is a compiler with a process, not a chatbot that hands you snippets. You start with a conversation: type, talk, or paste in a document. Remy asks clarifying questions (“Should admins delete records, or just archive them?”), then drafts the plan, a plain-language spec with light annotations that pin down the details. You read it, approve it, and adjust anything that’s off.
Then it compiles. Under the hood, Remy works like a team rather than a single model. Six specialist sub-agents split the job, all coordinated against the plan:
- Coding: writes the backend methods and application logic
- Design: visual identity, typography, and layout, drawing on a curated font catalog and inspiration library
- Architecture: decides how the pieces of the app fit together
- QA: drives a real browser, clicking through the app to confirm the flows actually work (and recording a walkthrough as it goes)
- Roadmap: maintains a living roadmap of the build as a first-class artifact, not a buried chat log
- Research: gathers the context and references the build needs
That division of labor is why the result is a coherent application, not a pile of files.
The plan is written in MSFM (MindStudio-Flavored Markdown), plain Markdown with a few annotations, so it stays readable and portable. (Want the exact format? The MSFM walkthrough goes deep.) The whole loop, from first message to live app, takes anywhere from ten minutes to an hour.
What does Remy build?
One compile produces a complete, production-ready app, not a starter kit you finish by hand:
- Backend: TypeScript methods with typed inputs and validation, plus any npm package
- Database: serverless SQL with typed schemas and automatic migrations
- Auth: email/SMS verification codes, sessions, and role-based access, opt-in
- Frontend: a Vite + React app (or another framework you name), CDN-hosted and mobile-responsive
- Deployment: git-native. Push to main and it builds, tests, and deploys, with atomic releases and rollback
- Interfaces: the same backend answers through web, REST API, Discord, Telegram, webhooks, cron, and MCP
- Docs: API docs and a deployment guide, generated alongside
The verification codes actually send, the SQL actually runs, and sessions actually persist. That’s the line between a prototype and a product. (See how one backend projects to eight interfaces.)
How is Remy different from coding agents and app builders?
Remy is a product agent, a different category from the two kinds of AI tools people usually compare it to. The quickest way to see it:
| Coding agents (Cursor, Claude Code) | App builders (Lovable, Bolt, v0) | Remy (product agent) | |
|---|---|---|---|
| You start with | an existing codebase | a prompt | an idea or a plan |
| The AI | edits code line by line | generates a frontend | compiles the whole stack |
| What you keep | the code | a chat log + UI | the plan |
| Backend & database | you write them | usually none | built in |
| Best for | refining existing code | UI demos | shipping a new app end-to-end |
Coding agents are pair programmers for code you already have. App builders make a good-looking screen from a prompt, usually with no real backend behind it. Remy works a level up: you describe the product, and it compiles the entire stack. For the full breakdown, see product agent vs coding agent; for a head-to-head, Remy vs Lovable.
What can you build with Remy, and who is it for?
Remy is at its best on full-stack web apps where the value is in the workflow: internal tools, vertical SaaS, CRMs, approval flows, dashboards, content tools, and AI-native apps. It fits a wide range of people: product managers who can describe data and rules without writing code, founders who need a real MVP (not a mockup), operators who need an internal tool this week, and developers who’d rather describe an app and drop into the code only when they want to.
A few workloads sit outside its sweet spot, and the move there is to match the tool to the job: native mobile (Remy ships mobile-responsive web; use a native toolchain for offline-first apps), real-time multiplayer or sub-100ms collaboration (use a realtime-first stack), extreme write-heavy workloads (pair it with an external Postgres), and enterprise SSO/SAML (on the roadmap; OAuth and verification codes work today). For the apps most teams actually need to ship, Remy owns the whole stack, precisely because it isn’t trying to be everything.
What is Remy built on?
Remy is open source. The agent, the SDKs, and the local dev tooling are public under github.com/mindstudio-ai. You can read it, fork it, or run it locally, and the code for each app you build lives in a git repo you own.
It runs on a production platform hardened by years of real enterprise traffic, which is why a Remy app gets so much without setup: 200+ AI models, 1,000+ integrations, managed databases, auth, and deployment, all out of the box. Remy is the product-agent layer on top: the platform handles the hard infrastructure, and Remy compiles your plan into a working app. And these aren’t demos. The Debut gallery breaks down real apps people have shipped and use.
FAQ
How do I get access to Remy? Remy is in open alpha. Sign up and start building at goremy.ai.
What does Remy cost? Remy is $99/month, or $79/month with annual billing (about two months free), after a 7-day free trial. You get the full product, with no feature limits during the trial. On top of that, inference (the raw AI model usage) is pass-through at provider rates with no markup: fund a balance or bring your own provider keys. A typical app costs around $100 in inference to build, and iteration costs less because only what changed recompiles.
Can I edit the generated code? Yes. It’s standard TypeScript in a git repo you own. You can edit it directly, or update the plan and recompile. Recompiling is the recommended path, so your changes survive the next build.
What about auth and security? Auth is opt-in and built into every app: email or SMS verification codes, sessions, and role-based access enforced at the method level. Secrets are encrypted at rest, with separate dev and prod values.
What models does Remy use? It’s model-agnostic and uses the best model for each job: Claude Opus for the core agent, Sonnet for the sub-agents, plus image models. As models improve, recompiling the same plan produces better output without any rewrite.
Can I deploy to my own infrastructure? The code is portable (TypeScript, React, SQL), but the runtime and database are part of the platform, so apps deploy to managed infrastructure by default. Your plan and code are yours to take.
Can Remy build a native mobile app? Not native. Mobile-responsive web apps work well, but native iOS/Android isn’t supported.
Is Remy production-ready? The apps run on a platform that’s carried real enterprise production traffic for years, at SOC 2 Type II and GDPR scale. The product is in alpha, so expect bugs and changes, but the runtime is production-grade.
What’s the learning curve? If you can describe what an app should do, you can build one, with no backend experience required. Remy handles the infrastructure.
The bottom line
Most AI tools hand you code or a screen and leave the hard part to you: the backend, the database, the deploy. Remy hands you the whole app, built from a plan you can read and change, and gets better every time you recompile as the underlying models improve.
Remy is a product agent that compiles annotated markdown into a full-stack app (backend, database, frontend, auth, tests, and deployment) in a single step. See goremy.ai.
