// the find
GoogleCloudPlatform/agent-starter-pack
Ship AI Agents to Google Cloud in minutes, not months. Production-ready templates with built-in CI/CD, evaluation, and observability.
A Python CLI that scaffolds production-ready GenAI agent projects for Google Cloud — Terraform, CI/CD, evaluation harnesses, and observability wired up from the start. It's for teams building on Vertex AI/ADK who want to skip the weeks of infra plumbing between a working prototype and a deployed agent. Templates cover ReAct, RAG, multi-agent, and Live API patterns in Python, Java, Go, and TypeScript.
Genuinely broad template coverage — ADK, LangGraph, RAG with both Vertex AI Search and Vector Search backends, A2A protocol, and a real-time multimodal agent, not just one happy-path demo. The `enhance` command retrofits infra/CI/CD onto an existing agent directory instead of forcing a greenfield-only workflow. Terraform and GitHub Actions/Cloud Build pipelines are generated per-environment (dev/staging/prod), which is more than most starter kits bother with.
The repo is in maintenance mode as of this snapshot — Google states no new features, templates, or deployment targets will land, and is steering new users to a separate tool (agents-cli) instead. Anyone adopting this today is building on something the maintainers have already deprecated. It's Google Cloud only: Terraform modules, Vertex AI eval, and Cloud Run/Agent Engine deployment targets mean zero portability if you're not already committed to GCP. Getting past the `uvx create` toy example requires gcloud SDK, Terraform, and Make installed and configured — the 1-minute pitch undersells the real setup cost for production use.