Most AI app builders assume you are starting from scratch. But if you already have a working codebase, the question is different: how do you import a project into an AI app builder without losing your setup, your Git history, or your sanity?
Dyad is built for this. It is a local, open-source desktop app that runs on Mac, Windows, and Linux. You can point it at an existing folder or pull a repo from GitHub, and the AI starts working with your code immediately. No uploads, no copying files to a cloud service, no proprietary project format.
Why you would import a project into an AI app builder
There are a few common reasons:
- You built a prototype in another tool (Lovable, Bolt, v0) and want to keep going without the subscription cost or feature limits.
- You have a side project on GitHub and want AI help adding features, fixing bugs, or refactoring.
- You want to use AI on a work project locally, without sending your source code to a third-party cloud.
- You are migrating from another AI coding tool and want to keep your existing codebase intact.
In all of these cases, the goal is the same: bring your code into a tool that understands it and can edit it, without forcing you to start over.
How to import a project into Dyad
Dyad supports two import paths: local folders and GitHub repositories.
Import from a local folder
- Open Dyad and click the + button to create a new app.
- Select Import App.
- Choose the folder containing your project.
- Dyad reads the project structure and sets up a local dev environment.
By default, Dyad runs npm run dev to start the local development server. If your project uses a different start command, you can configure custom install and start commands during import.
Starting with v0.33.0, Dyad supports in-place importing. This means your project stays in its original directory. Dyad does not copy it to ~/dyad-apps/ or any other managed folder. If you have a monorepo, a specific disk layout, or just prefer to keep things where they are, in-place import handles that.
Import from GitHub
If your project lives on GitHub, you can import it directly.
- Connect your GitHub account in Dyad's settings.
- Click + and select Import from GitHub.
- Browse or search your repositories and select the one you want.
- Dyad clones the repo locally and sets it up.
This was added in v0.25.0, and it works with both public and private repositories. After importing, the GitHub remote stays connected, so you can push changes, pull updates, and manage branches from within Dyad.
Auto-generated AI_RULES.md for imported projects
When you import a project into an AI app builder, the AI needs context about your codebase. What framework are you using? What conventions should it follow? Where do components live?
Dyad handles this with a file called AI_RULES.md in the root of your project. This file contains instructions that guide the AI when making changes. It describes your tech stack, architecture, and any coding standards you want enforced.
If your project does not already have an AI_RULES.md, Dyad generates one automatically during import. It inspects your project structure, detects your framework and dependencies, and writes a set of rules that match your setup.
You can edit AI_RULES.md at any time. A few examples of rules worth adding:
- "Use Tailwind utility classes instead of custom CSS."
- "All database schema changes must be backwards compatible."
- "Place new API routes in the
src/api/directory."
The AI reads this file before every interaction, so it stays aligned with your project's conventions as the codebase grows.
Cross-app references
One useful pattern when working with imported projects: cross-app references. If you have multiple apps in Dyad, you can type @ in the chat input and reference another app. The AI then reads the referenced app's code alongside your current project.
This is helpful when you want to:
- Port a feature from a prototype into your main codebase.
- Build a new app that follows patterns established in an existing one.
- Split a large app into smaller pieces and still let the AI see the full picture.
For Dyad Pro members, Smart Context works across referenced apps too. It automatically selects the most relevant files from both the current app and any referenced apps, so you do not burn tokens sending entire codebases.
What to keep in mind
Importing an existing project into an AI app builder is not always frictionless. A few things to watch for:
JavaScript and Node.js only (for now)
Dyad currently supports Node.js-based projects. If your app uses PHP, Ruby, Python, or another non-JavaScript runtime, importing will not work yet. Most modern web frameworks (React, Next.js, Vue, Svelte, Vite-based projects) work fine.
Your project needs a dev script
Dyad runs npm run dev to start the preview server. If your package.json does not include a dev script, you need to add one, or configure a custom start command during import.
Large codebases and token limits
If your imported project is large, you may run into model context limits. Dyad offers manual context management (glob-style file patterns) and Smart Context (Pro) to control which files the AI sees. For very large projects, consider focusing the AI on specific directories rather than the entire codebase.
Standard code, no lock-in
A critical difference between Dyad and cloud-based AI app builders: your imported project stays as standard code. Dyad does not wrap it in a proprietary format, inject a custom runtime, or require a platform-specific deployment target.
After importing, your project is still a normal Git repository. You can open it in VS Code, run it from the terminal, deploy it to Vercel or any other host, or stop using Dyad entirely. The code is yours.
Dyad is open-source (MIT license, with FSL 1.1 for pro features) and supports multiple AI providers, including OpenAI, Anthropic, Google, and local models via Ollama or LM Studio. You bring your own API keys, so there is no per-message pricing or credit system to worry about.
Getting started
If you have an existing project and want to try AI-assisted development, the fastest path is to download Dyad, import your project, and start chatting with the AI about your code. The import process takes under a minute for most projects.
For projects already on GitHub, connecting your account and importing a repo is a few clicks. For local folders, just point Dyad at the directory.
Either way, the result is the same: your existing codebase, enhanced with AI that understands your project's structure and follows your rules.
