Dyad review

Dyad is a local, open-source desktop app builder that uses AI to turn prompts into web apps on your computer. You choose the model, rather than being locked into one provider: the product supports options including OpenAI, Anthropic, Google, Ollama and LM Studio. The Free plan also lets you bring your own API key.

The practical difference from a hosted app builder is that Dyad keeps project files on your machine and gives you code you can edit, export or open in VS Code. You can import repositories, connect Supabase or Neon for database and authentication work, scan for security issues, and deploy through GitHub and Vercel or your own cloud.

That makes Dyad a fit for people who want AI help without handing the project entirely to a closed platform. It is still a builder, not a promise that every app will be finished without developer input. The code remains yours to work on.

Dyad’s clearest advantage is control. It combines a local project folder and editable code with a choice of AI models, so you are not required to build around one model provider or keep the app inside a vendor’s hosted environment. For a switch from a closed builder, that makes it easier to keep working in familiar coding tools or move the code elsewhere.

Other useful parts of the workflow, according to the product details, include:

• Bring in an existing repository, edit files in VS Code and export runnable code.
• Connect Supabase or Neon for schemas, migrations and authentication.
• Run a security scan for bugs, injection issues and misconfigurations.
• Deploy to GitHub and Vercel, or use your own cloud.

The entry point is also genuinely usable without a subscription: Dyad Free costs $0, includes 20 Basic Agent messages per day, supports your own API key and requires no sign-up, according to the pricing page. Reddit users describe building projects locally, including a poker game; that is anecdotal feedback, not evidence that every app will be production-ready.

The free tier has a firm usage ceiling: 20 Basic Agent messages per day. The pricing page lists Dyad Pro at $20 per month for 200 AI credits and Dyad Max at $79 per month for 900 credits. Those are monthly credit allowances, not message counts, so they are not directly comparable with the Free plan’s daily messages. Annual prices are not published, and the price for reloading credits is not specified, though Dyad says reloads are available at the same price. Token orchestration is listed at 1.5 credits per million tokens.

Local-first also means project files live on your machine, rather than automatically making the setup a shared, synced workspace. In a Reddit thread, Dyad’s founder said collaboration features for syncing GitHub changes were in progress, while broader syncing for settings and chats was something to consider. The founder also noted that Supabase database changes are not easy to undo; a Neon-based portal template supports database undo. If shared workflows or easy database rollbacks are requirements, check their current status before switching.

Dyad Free costs $0 and includes 20 Basic Agent messages per day, local app-building, macOS and Windows downloads, and bring-your-own API key access. Dyad Pro is $20 per month for 200 AI credits, Pro modes, and Academy access. Dyad Max is $79 per month for 900 credits and prioritized office-hour access. That’s a $59 monthly jump from Pro to Max, for 700 more monthly credits plus the listed Max benefit.

The main catch is that free messages and paid credits are different units, so the pricing page doesn’t provide a direct usage comparison. Pro credit reloads are available at the same price, but the reload price isn’t specified. Dyad also lists a charge of 1.5 credits per million tokens for token orchestration. Prices are monthly; no annual price is published. The listed starting price is $20, though the free plan means you can start without paying.

Moving an existing project into Dyad can start with importing its repository; Dyad says you can edit in VS Code and export clean, runnable code. Its local-first setup keeps project files on your machine, so leaving need not mean abandoning the code in Dyad. The listed deployment options include GitHub, Vercel, or your own cloud. Dyad also connects to Supabase or Neon for database schemas, migrations, and authentication.

The trade-off is that Dyad isn’t a fit for teams that need built-in general syncing, and its paid plans cap monthly AI credits. Those are plausible reasons to switch away, but the published published material contains no Reddit posts, so we can’t attribute departure reasons to Reddit users. Before moving, check how your current app’s database, authentication, and deployment setup maps to Dyad’s integrations; the published information doesn’t specify an automatic migration path for those services.