Skip to main content
Apps are managed web applications. Each app gets managed hosting (deployments, domains) plus — for fullstack apps — a dedicated managed backend (PostgreSQL, auth, storage, edge functions). Naive provisions and owns the underlying infrastructure and credentials, so you never need any separate account.
The common app lifecycle has first-class endpoints, and everything else is reachable through the hosting proxy and backend proxy, which inject credentials and scope each call to your app’s own project.

Two Ways to Work

Apps are fully standalone — no AI orchestration required. Choose the mode that fits:

Direct Quickstart

naive apps templates lists every available template with its clone command.

App Types

Starter Templates

Templates live in the public repo usenaive/app-dev-templates, organized as {type}/{variant}: Every app’s create response (and naive apps show) includes a template block with the repo URL, path, and a ready-to-run cloneCommand. In orchestrated mode the same template is scaffolded into the engineer agent’s workspace automatically.

Creating an App

Returns the app ID, hosting project details, the default production domain (e.g., naive-studyai-abc123.vercel.app), and the template block. For fullstack apps, backend provisioning starts immediately in the background. Once the project is healthy, the backend URL and anon key are automatically injected as environment variables.

Deployment Workflow

Preview Deployment

The CLI packs your project (excluding node_modules, .next, .git) and uploads it through the Naive API. Inside agent containers, the same command deploys the agent’s workspace instead. Returns a unique preview URL — poll naive apps deployments <app-id> for build status.

Promote to Production

Waits for the deployment to be READY, then aliases it to the production domain. Zero-downtime switch.

Deployment History

Environment Variables

Secrets are stored encrypted by Naive and synced to the app’s environment variables (production → production; preview → preview + development):
Redeploy after changing secrets for them to take effect.
Provisioned automatically: NEXT_PUBLIC_APP_URL (all apps), NEXT_PUBLIC_SUPABASE_URL and NEXT_PUBLIC_SUPABASE_ANON_KEY (fullstack apps).

Domains

For {slug}.usenaive.ai: do not domains set-record or dig-and-wait 48h. A company dns_status=pending_verification is Resend email lag — ignore it for app attach (app_connect_status / verify-dns is the front door).

Backend Capabilities (Fullstack Only)

A fullstack app’s backend is exposed as four first-class primitives — usable for your own app and, via forUser(id), for each of your end-users’ apps:
In your app code, use @supabase/supabase-js with the auto-injected env vars for queries, auth, and storage.

Full Platform Access (Advanced)

The proxies expose the entire underlying hosting and backend platform APIs, scoped to your app’s own project:
See the hosting proxy and backend proxy references for scoping rules, and the official hosting / backend platform docs for the full operation catalog.

Retry Failed Provisioning

Re-creates the hosting project and/or backend if they were never linked (works even when the app shows active but its background backend provisioning failed).

Deleting an App

Deletes the hosting project, the backend (if fullstack), all secrets, domains, deployment records, and archives the associated engineer agent. Irreversible.

Agent Integration (Orchestrated Mode)

When the company has an agent container and the CEO creates a plan that includes apps, it:
  1. Creates apps with naive apps create — the starter template is scaffolded into a dedicated engineer agent’s workspace
  2. Creates tasks with explicit deployment instructions
  3. Workers build the code and run naive apps deploy + naive apps publish automatically
Apps created before the container exists work too — the engineer agent is registered as pending and orchestration picks it up later, while you can build and deploy directly the whole time.