Cezzis Cocktails

Part of the broader Cezzis.com digital experience for discovering and sharing cocktail recipes with a broad community of cocktail enthusiasts and aficionados.

This repository contains the Go-based MCP server for Cezzis.com cocktails. It exposes a streamable HTTP MCP interface that lets MCP clients search cocktail data, retrieve cocktail details, authenticate against Auth0, and submit user ratings through the Cezzis platform APIs.

The server is an integration layer, not the source of cocktail data. It registers MCP tools, forwards requests to the upstream Cocktails, AI Search, and Accounts APIs, stores authentication tokens in PostgreSQL per MCP session, and emits telemetry through OpenTelemetry.

- Search cocktails by free text.
- Retrieve full cocktail details by cocktail ID.
- Start and manage Auth0 device-flow authentication.
- Submit authenticated cocktail ratings.
- Expose health and MCP HTTP endpoints for local and deployed environments.

- Go 1.25.1
- Model Context Protocol over streamable HTTP viamark3labs/mcp-go
- OpenAPI-generated API clients for upstream Cezzis services
- Auth0 device authorization flow
- PostgreSQL for MCP session token storage
- OpenTelemetry and zerolog for observability
- Kubernetes manifests for local deployed environments under.iac/k8s
- Terraform for production Azure infrastructure under.iac/terraform

. ├── .iac/ │ ├── argocd/ # Argo CD manifests for cluster sync │ ├── k8s/ # Local Kubernetes deployment manifests │ └── terraform/ # Production Azure infrastructure ├── .vscode/ # IDE launch configuration ├── cocktails.mcp/ │ ├── http-client.env.json │ └── src/ │ ├── cmd/ # Application entry point │ └── internal/ │ ├── api/ # Generated API clients │ ├── auth/ # Auth0 flow and token handling │ ├── db/ # PostgreSQL connection and setup │ ├── mcpserver/ │ ├── middleware/ │ ├── repos/ │ ├── telemetry/ │ └── tools/ # MCP tool definitions and handlers ├── Dockerfile ├── makefile └── mcp.http

Tool execution requests rely on theMcp-Session-Idheader so the server can associate requests with an MCP session.

- Go 1.25.1+
- PostgreSQL
- Valid values for the upstream API hosts and subscription keys
- Auth0 settings if you want to use authenticated tools

Create a.envfile incocktails.mcp/srcwith the values your environment needs:

PORT=7999 ENV=loc COCKTAILS_API_HOST=https://your-host/cocktails COCKTAILS_API_XKEY=replace-me ACCOUNTS_API_HOST=https://your-host/accounts ACCOUNTS_API_XKEY=replace-me AISEARCH_API_HOST=https://your-host/search AISEARCH_API_XKEY=replace-me AUTH0_DOMAIN=your-tenant.us.auth0.com AUTH0_NATIVE_CLIENT_ID=replace-me AUTH0_ACCOUNTS_API_AUDIENCE=https://api.cezzis.com/accounts AUTH0_SCOPES=openid offline_access profile email read:owned-account write:owned-account CEZZIS_BASE_URL=https://www.cezzis.com POSTGRES_HOST=localhost POSTGRES_PORT=5432 POSTGRES_DB=cezzis_cocktails_mcp POSTGRES_USER=postgres POSTGRES_PASSWORD=replace-me POSTGRES_USE_TLS=false

make compile ./cocktails.mcp/dist/linux/cezzis-cocktails

The server listens on the configuredPORT. By default that is7999.

For IDE-based debugging, use the launch configuration in.vscode/launch.json. It runs the Go application fromcocktails.mcp/src/cmdwith local environment loading enabled.

The manifests in.iac/k8sare for a local deployed environment. They define the Deployment, Service, Ingress, ConfigMap, and ExternalSecret wiring used when running the app in a local cluster.

# loc app kubectl apply -f https://raw.githubusercontent.com/mtnvencenzo/cezzis-com-cocktails-mcp/refs/heads/main/.iac/argocd/cezzis-com-cocktails-mcp-loc.yaml # loc image updater kubectl apply -f https://raw.githubusercontent.com/mtnvencenzo/cezzis-com-cocktails-mcp/refs/heads/main/.iac/argocd/image-updater-loc.yaml # cloudsync app kubectl apply -f https://raw.githubusercontent.com/mtnvencenzo/cezzis-com-cocktails-mcp/refs/heads/main/.iac/argocd/cezzis-com-cocktails-mcp-cloudsync.yaml # cloudsync image updater kubectl apply -f https://raw.githubusercontent.com/mtnvencenzo/cezzis-com-cocktails-mcp/refs/heads/main/.iac/argocd/image-updater-cloudsync.yaml

Production infrastructure for this application is defined in.iac/terraform. In production, the MCP app is hosted in Azure Container Apps, sits behind Azure API Management, and is reached externally through Azure Front Door.

This project is proprietary software. All rights reserved. SeeLICENSEfor details.

Free search API for AI agents with 105 engines, 22 search profiles, and 94.3% SimpleQA accuracy. MCP server included.