Integrations

Choose the connection you needCopy Anchor Link

Pagiera runs inside your application. Framework adapters, backend services and data sources play different roles: PostgreSQL and Redis are required by the bundled backend; AI and MCP connections are optional. Your application supplies authentication. The integration catalog describes these connections, rather than a marketplace of installable plugins.

1. Connect the host and required servicesCopy Anchor Link

Start with an isolated development application, PostgreSQL database and Redis service. Install pagiera, import pagiera/full.css, then keep the service URLs in server-only configuration:

PAGIERA_POSTGRES_URL=postgresql://postgres:postgres@localhost:5432/pagiera
PAGIERA_REDIS_URL=redis://localhost:6379

Follow Next.js setup to mount the protected API, load the editor bootstrap and render the studio. The backend requires a Node.js runtime; a static export or Edge-only deployment cannot run these connections. React rendering alone does not provision a backend or authentication.

Verify: sign in, check /api/pagiera/health, create a page, save and reload. Then sign out and confirm editor operations are denied. Keep the database and Redis private. A failed health request is a service or authorization problem to diagnose before testing AI or data bindings.

2. Connect an API or CMSCopy Anchor Link

Any service exposing a suitable JSON endpoint can supply page data. This is an HTTP data-source connection; it does not imply a built-in SDK or account connector for every CMS.

1. Add the endpoint in the studio's data-source interface and inspect the response.
2. Use Request for an object or Repeat for a collection. Select a nested list path such as $.items when needed.
3. Bind child layers to returned fields. Pass route values with tokens such as {{params.slug}}.
4. Preview empty, failed and not-found responses before publishing.

Do not place permanent credentials in page URLs or headers: editors can see the document. Use an application-owned proxy that adds secrets on the server and validates requested resources. Private hosts are blocked unless the operator explicitly allows them. See source limits and failure behavior before connecting internal services.

3. Add optional AI and agent toolsCopy Anchor Link

For Luma, set OPENROUTER_API_KEY and OPENROUTER_MODEL on the server. Verify provider access and billing, then generate within a limited editing scope. Review the resulting layers and preview them before publishing. The editor's manual tools work without an AI key.

For MCP discovery, configure approved mcpServers and implement authorizeMcp. Connections can use stdio, HTTP or SSE; stdio also needs the executable installed in the host environment. With no authorization callback, discovery is denied. Listing or inspecting tools does not automatically execute them. See the AI and MCP guide for the configuration contract.

The CLI is a separate way for coding agents to inspect and update native documents. It uses the host API and its authorization rules; it is not a substitute for a working server integration.

Diagnose a connectionCopy Anchor Link

Use troubleshooting for the next diagnostic steps. Never copy production credentials into browser configuration or expose an unprotected editor to make a connection work.