#1★ TOP PICK
Dyad
A free desktop app builder with a Next.js and shadcn template, any model, and your code on your own disk.
93
OPEN SOURCEApache-2.0 (src/pro under FSL-1.1-ALv2)SELF-HOSTLOCAL-FIRST
Dyad is a local, open-source AI app builder that describes itself as an alternative to v0, Lovable and Bolt that runs on your own computer. You chat, it builds and previews the app, and every project is a regular Git repository in a folder on your machine (~/dyad-apps by default), so you can open it in VS Code whenever you like. Alongside its default React and Vite template it ships a Next.js template set up with shadcn and Tailwind, the stack v0 generates. You bring the model: Google Gemini, OpenAI, Anthropic, OpenRouter, a ChatGPT subscription, or local models through Ollama or LM Studio. Supabase and Neon connect for a database and auth, and its docs walk through deploying with GitHub and Vercel or any other host.
⌁ Dyad runs anywhere Node.js runs, and its FAQ suggests a cloud model such as Gemini or OpenRouter if your machine is not powerful. Local models through Ollama or LM Studio are optional; Dyad's own docs warn they are resource-intensive and that smaller models may struggle to follow its instructions.
Strengths
- +Each project is a normal Git repository on your own disk; open it in any editor, or keep building elsewhere
- +A Next.js template with shadcn and Tailwind, so the output looks like what v0 produces
- +Any model: Gemini, OpenAI, Anthropic, OpenRouter, a ChatGPT subscription, or local models via Ollama or LM Studio
- +Very active: v1.18.0 shipped on 5 October 2026, and the repository has about 21.7k GitHub stars
Trade-offs
- −A desktop app rather than a hosted service: you install it plus Node.js, and macOS and Windows are the supported platforms; Linux is experimental and does not auto-update
- −JavaScript apps only, per its FAQ: web apps and hybrid mobile apps, but no native Swift or Kotlin
- −The Pro features in the repository's src/pro folder are fair-source (FSL-1.1), not open source, and the free agent mode is limited to 20 messages a day
- −Dyad does not host your app; you deploy through GitHub to Vercel or another host and manage that account yourself
Free and open source; bring your own API key or a ChatGPT subscription, or run a local model. The free Basic Agent is listed at 20 messages a day. Optional Dyad Pro is $20/month with 200 AI credits, and Dyad Max $79/month with 900. Verified on dyad.sh/pro 2026-10-06. #2
Cline
For developers: the same prompt-to-code loop, inside your own repository and editor.
92
OPEN SOURCEApache-2.0SELF-HOSTLOCAL-FIRST
If you mainly use v0 to generate Next.js and shadcn/ui code that ends up in a repository you already have, an open coding agent can do that job in place. Cline (Apache-2.0) runs as a VS Code extension, a JetBrains plugin, a CLI and a desktop app for macOS and Windows. It reads your project, plans before it acts, shows every edit as a diff you can review, asks before running commands, and keeps checkpoints you can roll back to. It works with Anthropic, OpenAI, Google, OpenRouter, Vercel AI Gateway, Bedrock, any OpenAI-compatible endpoint, or local models through Ollama or LM Studio. It is a coding agent, not an app builder: there is no hosted preview or one-click deploy, so it suits people comfortable running a dev server themselves.
⌁ Cline runs anywhere VS Code or Node.js runs; the hardware question is only about the model. A cloud provider needs nothing special, while local inference through Ollama or LM Studio needs a capable GPU or a machine with plenty of unified memory.
Strengths
- +Works on the repository you already have, so generated components land beside your real code
- +Every edit shows as a reviewable diff, with checkpoints to undo the agent's work
- +Any provider, or fully local with Ollama or LM Studio
- +Large and active: about 70k GitHub stars, with commits daily
Trade-offs
- −No live preview, hosting or deploy button of its own; you run the dev server and deploy yourself
- −Assumes you are comfortable in a code editor; it is not built for people who have never run a project locally
- −The JetBrains plugin is not open source; only the VS Code extension, CLI, SDK and desktop app are in the public repository
- −Per-token billing on a frontier model can cost more than a flat plan for heavy daily use
Free for individual developers (Apache-2.0); you pay only for inference, with your own keys or Cline's usage-billed provider. Optional ClinePass is $9.99/month for open coding models; Enterprise is custom. Verified on cline.bot/pricing and docs.cline.bot 2026-10-06. #3
bolt.diy
The open-source edition of Bolt.new: a browser-based app builder you run yourself, with any model.
78
OPEN SOURCEMITSELF-HOSTLOCAL-FIRST
bolt.diy is the official open-source version of Bolt.new, published under StackBlitz Labs and maintained as a community project. You run it on your own machine (Node.js and pnpm, or Docker) and get a browser-based builder with chat, an editor, a terminal and a live preview of Node.js-based web apps, using whichever of its more than 20 providers you choose, including Ollama and LM Studio for local models. Projects download as a ZIP, connect to Git, and deploy to Netlify, Vercel or GitHub Pages. One catch is specific to it: apps run inside StackBlitz's WebContainer API, which StackBlitz licenses separately and which needs a commercial licence for production use in a for-profit setting.
⌁ bolt.diy itself is light: a local Node.js server or Docker container serving a browser UI. With a cloud model it runs on modest hardware. Local models through Ollama or LM Studio need a capable GPU or plenty of unified memory, and the README lists better performance for smaller models as planned work.
Strengths
- +Chat, editor, terminal and live preview in one browser tab, running on infrastructure you control
- +More than 20 providers, including Anthropic, OpenAI, Google, OpenRouter, Ollama and LM Studio
- +Download any project as a ZIP, connect Git, and deploy to Netlify, Vercel or GitHub Pages
- +An active community project: about 19.9k GitHub stars, with commits landing this week
Trade-offs
- −Its runtime, the WebContainer API, is proprietary; StackBlitz requires a commercial licence for production use in a for-profit setting
- −No one-click install: you run it from source with Node.js 22.12+ and pnpm, or Docker; the last tagged release and desktop build (v1.0.0) date from May 2025
- −Its README lists the move from single model calls to an agent-based system as still in progress
- −No hosting or database of its own: you deploy to Netlify, Vercel or GitHub Pages and connect Supabase yourself
Free and open source (MIT); you pay your model provider, or nothing beyond hardware with a local model. StackBlitz requires a separate commercial WebContainer API licence for production use in a for-profit setting; prototypes and proofs of concept do not need one. #4
Google AI Studio
Google's free, hosted prompt-to-app builder, if a Gemini-only tool and Google's free-tier terms suit you.
25
SOURCE-AVAILABLEProprietary (hosted service)
Build mode in Google AI Studio turns a prompt into a full-stack web app (a React front end and a Node.js server by default) or a native Android app in Kotlin and Jetpack Compose, with a live preview in the browser. It can provision Firebase Firestore and Firebase Authentication, sync two ways with GitHub, export the project as a ZIP, and deploy to Cloud Run. Google's billing page says AI Studio usage is free of charge unless you link a paid API key, which makes it the lowest-cost hosted option here. Two trade-offs come with that: it is built around Google's Gemini models, and on the free tier Google's terms say your prompts and outputs may be used to improve its products and may be read by human reviewers.
Strengths
- +No subscription and no setup: free in the browser unless you link a paid key
- +Full-stack by default (React plus a Node.js server), and native Android apps as well
- +Two-way GitHub sync and ZIP export; a downloaded app runs on any host once you set GEMINI_API_KEY
- +Firebase Firestore and Google sign-in can be provisioned for you
Trade-offs
- −Built around Gemini; Google's Build mode guide describes no other model provider
- −On the free tier Google may use your prompts and outputs to improve its products, human reviewers may read them, and its terms say not to submit sensitive or confidential information
- −A hosted Google service you cannot self-host; deploying from AI Studio means Cloud Run, billed by usage
- −When you share an app, its Gemini API calls count toward your own usage limits
Free to use unless you link a paid API key; deploying to Cloud Run is billed by Google Cloud based on usage, and apps that call paid models can incur costs. Verified on ai.google.dev 2026-10-06.