ProjectBuild a Permission-Aware Knowledge Hub as an MCP Server: the explanation · Module 20: Deploying to Google Cloud with CI/CD · part 22 of 22
Part 22 · Module 20: Deploying to Google Cloud with CI/CD
Module 20: Deploying to Google Cloud with CI/CD
25 min read·
About 45 minutes · Needs: Module 19 (the containers and the local stack), Module 18 (X-Forwarded-For, rate limits and demo accounts), Module 12 (OAuth and the token audience), Module 13 (elicitation in create_issue) · Code: deploy/terraform/*.tf, deploy/deploy.sh, deploy/cloudbuild.yaml, deploy/README.md, server/wayfinder/core/proxy.py, console/nginx.conf.template, .github/workflows/ci.yml, .github/workflows/deploy.yml
Where this fits
The parts this module puts in the cloud (blue): the deploy and CI/CD, the same containers, Postgres as Cloud SQL, nginx as the one public front door, and the proxy-secret guard.
Module 19 ran the system on one laptop. This module puts the same two images on Google Cloud, so that AI clients anywhere, including Claude's custom connectors, can reach Wayfinder over HTTPS, and every commit that passes the tests is deployed without anyone holding a cloud key.
Read this first: none of this has run. The Terraform was written and documented but never validated (Terraform was not installed where it was written) and never applied to a real Google Cloud project, and the nginx change for Cloud Run has never run in a container. Every statement below about Cloud Run, Cloud SQL and the pipelines is design, not observation.
The rest of this project is yours to keep
This project is bought on its own, once, and stays readable afterwards, including the parts added to it later.