DOCUMENTATION

Source and self-hosting

Cookbook brings portable recipes, brand foundations, and reviewed deliverables together in a Next.js application. The GitHub repository contains the recipes, application, agent instructions, database migrations, and integration code.

Source and licensing

The marketing skill library is MIT licensed; see content/skills/LICENSE in the repository. This applies to content/skills and the generated skills distribution, not the application or hosted service. No application license is declared. Hosted plan access and source licensing are separate questions.

What an installation needs

The application uses Node.js 22 or newer, pnpm, and Supabase for authentication and saved work. Supabase may be hosted or local. The repository includes a Dockerfile and Compose configuration for the application; Compose does not start a Supabase database for you.

Recipes live in content/skills/, agent instructions in content/agents/cookbook/, integration services in lib/integrations/, and the connector in packages/cookbook-mcp/. Database migrations live in supabase/migrations/.

Configure the application

  1. From your local checkout, copy .env.example to .env.local.
  2. Set NEXT_PUBLIC_SUPABASE_URL and NEXT_PUBLIC_SUPABASE_PUBLISHABLE_KEY for your Supabase instance. Keep SUPABASE_SERVICE_ROLE_KEY server-side only.
  3. Apply the repository’s migrations in order to your own database. Set NEXT_PUBLIC_APP_URL to the application’s exact origin and configure Supabase authentication URLs for that origin.
  4. Set AI_GATEWAY_API_KEY for server-side model calls and review COOKBOOK_MODEL. Keep model credentials out of browser variables and source control.

For local development, run corepack enable, pnpm install --frozen-lockfile, and pnpm dev from the repository root. For the included container setup, run docker compose up --build; it reads .env.local and serves the application on port 3000. These start the application; database and model configuration are still required for saved, AI-assisted work. Configure and verify the background workflow runtime separately for your hosting environment; REST and MCP data access do not require model generation.

Enable foundations and connections

After applying migrations and verifying the setup, enable COOKBOOK_V2_ENABLED=true and COOKBOOK_INTEGRATIONS_ENABLED=true on the server. Set COOKBOOK_DEPLOYMENT=self-hosted for a self-hosted deployment. This disables commercial gates while retaining authentication, project permissions, and resource safeguards; it is not a source license.

Create a connection in your installation’s Settings page and follow the connection guide, using your installation URL. The local bridge needs a Cookbook connection key, not database administrator credentials. Hosted and self-hosted installations do not automatically share accounts, keys, or saved work.

Model support

The application’s server-side model calls use the configured AI Gateway. Provider-specific BYOK routing is not enabled simply by setting the optional OpenAI, Anthropic, or Google variables in the example environment file. That routing is future work. Portable recipes can be used independently in your chosen AI app, and connected apps use their own subscription or credentials.

Operate your installation

Use HTTPS for remote access, protect server secrets, and keep database backups with a tested restore process. Review migrations before upgrades and verify sign-in, foundation acceptance, draft saving, and connection revocation afterward. For container deployments, verify public Supabase configuration in the built application as well as runtime secrets before making the service available.