Get the Claude-on-GCP Operator Playbook
Google quietly turned Claude into a cloud-native team. Claude on Vertex AI — Opus 4.6, Sonnet 4.6, Haiku 4.5 — billed through your GCP account, $300 in new-account credits, plus startup/free-trial/committed-use credits all apply. This is the operator playbook from the Code with Claude 2026 SF live build (Ivan Nardini, Google Cloud DevRel) — zero to deployed feedback app on Cloud Run in 30 minutes, five roles, one developer. PAGE 01 THE DEAL — Vertex AI endpoint config (ANTHROPIC_VERTEX_PROJECT_ID, CLOUD_ML_REGION, claude --vertex), $300 burn, GCP account = the wallet. PAGE 02 THE 5-ROLE TEMPLATE — PM Claude (one-line brief → spec + acceptance + user stories), Designer Claude (UI layout + Tailwind + component tree), Engineer Claude (backend + frontend + schema), SRE Claude (Cloud Run config + env + health checks), QA Claude (writes + runs tests). PAGE 03 THE SUBAGENT CLAUDE.md TEMPLATE — paste-ready 5-subagent definitions with focused contexts (no pollution). PAGE 04 GCP MCP INSTALL — claude mcp add gcp / github / slack one-liners plus what each unlocks (Cloud Run, SQL, Logging, deploys, PRs, channel notifications). PAGE 05 THE 3 SKILLS — deploy-convention.md (Cloud Run defaults, region us-east5, secrets via Secret Manager), test-policy.md (real DB, real Cloud Run, every endpoint tested), naming-rules.md (snake_case tables, kebab-case services, PascalCase components). PAGE 06 THE $300 CREDIT-BURN CALCULATOR — how many small internal tools ship before credits run out (~50-70 at current Sonnet 4.6 pricing). PAGE 07 the 4 paid Hyperautomation Labs guides. The operator move: pick your next internal tool, fresh GCP project, run Claude Code with Vertex AI as the endpoint, spawn the 5 subagents, install the GCP MCP server, load the 3 skills, then prompt "build it, deploy it, give me the URL."
Free. No spam. Unsubscribe anytime.