The IDE for product managers.
mvpfy is a free, open-source desktop app that runs your company's existing product locally — describe a change in plain language, watch it live, and ship it as a pull request your engineers review. Like Lovable or Replit, but for the product you already have, on your own machine.
MIT-licensed · Windows & Linux builds on the releases page · bring your own agent (claude, optionally codex)
Describe → see it running → ship the PR
mvpfy manufactures the dev environment engineers take for granted, then hands the change loop to the person who owns the product.
Add your repos
Paste GitHub/GitLab URLs or a local folder — multi-repo products clone into one workspace. One click bootstraps everything into Docker: run config, a live .env, demo accounts, and open-source stand-ins (Postgres, Redis, MailHog) for anything missing.
See it running
The real product, live at localhost, inside the app — with demo logins surfaced, editable environment variables, streaming container logs, and a full VS Code editor one tab away. If a run fails, mvpfy diagnoses it in plain English and fixes it.
Ship as a pull request
Type a change in plain language — or implement a Feature1 user story with one click. The agent writes the code, runs the tests, and opens a PR on your real remote. Engineers review; nothing deploys behind their backs.
Everything between an idea and a reviewed PR
Live preview + VS Code built in
The running app and an embedded code-server editor live in tabs next to the overview — one window from idea to code.
Plain-language changes
“Add an env var”, “move the app to port 5000”, “add a PDF export button” — the agent applies the smallest safe change and reports back in one sentence.
Self-healing runs
When a bootstrap or start fails, mvpfy reads the log, explains what broke in plain English, fixes the workspace, and offers one-click retry.
Demo logins & env editor
Credentials to get into the app and editable environment variables sit on the overview — save, restart, applied.
Feature1 stories → PRs
Connected to Feature1, every user story gets an Implement button: acceptance criteria in, tested pull request out, backlog updated.
Local, private, MIT
Repos clone to your machine, the product runs in local Docker, and mvpfy has no server. Agents run under your own Claude Code or Codex account.
mvpfy vs Cursor vs Lovable & Replit
| mvpfy | Cursor / Windsurf | Lovable / Replit | |
|---|---|---|---|
| Built for | Product managers | Engineers | Builders of new apps |
| Unit of work | A product change or user story | A file / a diff | A prompt |
| Codebase | Your existing product | Your existing product | New, generated |
| Where it runs | Locally, in Docker | Locally | Their cloud |
| Output | A running app + a pull request | Reviewed diffs | A deployed app |
| Who reviews | Your engineers, via the PR | The engineer, inline | Nobody / you |
In short: Cursor makes engineers faster. Lovable builds new things. mvpfy makes PMs shippers on the product that already exists.
Questions, answered plainly
What is mvpfy?
mvpfy is a free, open-source desktop IDE for product managers. It clones your company's existing repositories, runs the product locally in Docker with a live preview and an embedded VS Code editor, and lets you describe changes in plain language that an AI coding agent applies and ships as pull requests for your engineers to review.
Is mvpfy free?
Yes. mvpfy is MIT-licensed and free. It is a thin wrapper over AI coding agents you bring yourself — Claude Code (required) and optionally Codex CLI — so agent runs use your own Anthropic or OpenAI subscription.
How is mvpfy different from Cursor or Windsurf?
Cursor and Windsurf are code-first editors that make engineers faster, one reviewed diff at a time. mvpfy is product-first: the unit of work is a change to the product, described in plain language or picked from a user story. The PM never reads a diff — the output is a running app to look at and a pull request for engineers to review.
How is mvpfy different from Lovable or Replit?
Lovable and Replit excel at building new apps in their cloud. mvpfy runs the same describe-see-ship loop on the product your company already has: your real repositories, cloned locally, running in Docker on your machine, with changes exiting through a pull request your engineers review rather than a deploy button.
Does my code leave my machine?
The product runs entirely locally: repositories are cloned to your machine and the app runs in local Docker containers. When you ask the AI agent to make a change, the agent sends relevant code to its model API (Anthropic or OpenAI) under your own account — the same as using Claude Code or Codex directly. mvpfy itself has no server and collects nothing.
What do I need to run mvpfy?
git, the GitHub CLI (gh), Docker Desktop, and Claude Code, signed in. Codex CLI is optional and only needed if you select it as your agent. mvpfy checks all of this on first launch and guides sign-in from Settings. macOS is the primary platform; Linux is supported and Windows is an experimental beta.
Can non-technical product managers use mvpfy?
The core loop is built for non-engineers: one-click bootstrap, a running app with demo logins surfaced, plain-language change requests, and self-healing runs that explain failures in plain English and fix them. The initial tool setup (installing git, Docker, and Claude Code) still benefits from an engineer's help — after that, the day-to-day loop needs no terminal.
Does mvpfy work with my existing repositories?
Yes — that is the point. Add one or more GitHub or GitLab URLs, or a local folder. Multi-repo products (frontend + backend + services) are cloned into one workspace and run together with a single generated Docker Compose file. Missing infrastructure is filled in with open-source stand-ins like Postgres, Redis, and MailHog.
Built in the open — and built the way it works
mvpfy is MIT-licensed, with CI, tests, and a documented architecture. It was built the way mvpfy itself works: a product owner directing coding agents. The repo is the demo. Issues and pull requests welcome — from humans and agents alike.
git clone https://github.com/feature1-ai/mvpfy.ai.git
cd mvpfy.ai && npm install
npm run dev
Ship your next change yourself.
Free, open source, and local. Your engineers still get the last word — in the PR.