Help Center

Brain Mode uses the Conductor agent to analyze your request, create an execution plan, and automatically delegate tasks to specialized agents like Senior Dev and Bug Hunter. Full guide coming soon.

Pro includes unlimited Brain workflows with premium models and token-based billing. BYOK (Bring Your Own Key) lets you use your OpenRouter API key and pay only for actual AI usage. Detailed comparison coming soon.

Yes. All code runs client-side in WebContainer, and API keys are encrypted with AES-256. Nothing leaves your browser except API calls. Full security documentation coming soon.

Go to Settings → Account and choose either an authenticator app (Google Authenticator, Authy, 1Password) or an emailed 6-digit code. It's optional and off by default. You'll get one-time backup codes at setup in case you ever lose access to your app or inbox.

Yes. Connect your own Supabase project and ForgeLab detects the data model your app needs, provisions the tables with row-level security and an auth trigger in one click, and wires the generated React, Vue, Svelte, or Next.js code to real database and Supabase Auth calls instead of mock data.

Yes, you choose the mode when reviewing the plan. Static export keeps client-side auth and stays deployable through ForgeLab's own Publish button. Full backend (SSR) adds real server-side session handling via middleware and works completely in Preview, but Publish is disabled for it, since it needs a real server, not a static host, so you deploy it yourself to a Node-capable host like Vercel to go live.

AI model requests in ForgeLab are routed through OpenRouter, the third-party service that connects us to the underlying AI providers. If OpenRouter or one of the model providers has a temporary outage, rate limit, or slowdown on their end, that shows up as an error in ForgeLab too, even though it isn't something wrong with ForgeLab itself. These issues are almost always short-lived, so waiting a moment and retrying usually resolves it.

Brain Mode's own Bug Hunter agent auto-reviews and retries fixes, but on a genuinely hard error it can get stuck without resolving it. Rather than repeating the whole multi-agent Brain workflow again (which re-runs every agent and costs a lot of tokens for one stuck file), switch off Brain Mode, use Single Chat, pick a stronger model from the Model dropdown, and paste the exact error from the Terminal or Preview and ask it to fix that specific issue directly. One strong model focused on one concrete error is faster and far cheaper than re-running the full team. Once it's fixed, run npm install and npm run dev in the Terminal tab if Preview doesn't refresh on its own.