agents-cli

agents-cli is an opinionated command center for building production-ready Google Cloud AI agents with the Agent Development Kit (ADK). It combines a polished Python CLI, installable coding-agent skills, project templates, evaluation workflows, deployment stubs, and a documentation site into one repeatable lifecycle.

The project is designed around a simple premise: agent development needs the same rigor as modern application delivery. Scaffold quickly, code with reusable agent skills, evaluate behavior, plan infrastructure, deploy intentionally, publish for discovery, and observe what happens in production.

Cloud integrations are intentionally stubbed in this initial implementation. Commands print the gcloud, Terraform, Docker, ADK, or registration action they would run and return success, so the workflow can be exercised immediately without credentials or infrastructure.

Why it matters

• Agent lifecycle in one binary: create, run, lint, evaluate, deploy, publish, and inspect from agents-cli.
• ADK-first project templates: standard ADK, Agent-to-Agent protocol, and agentic RAG scaffolds.
• Evaluation built in: JSON evalsets, trajectory matching, threshold configs, and score diffs.
• Cloud-ready without cloud lock-in during development: Terraform, gcloud, Docker, ADK, and Gemini Enterprise operations are stubbed until teams are ready to wire credentials.
• Skills for coding agents: seven markdown skills teach AI coding agents how to build, evaluate, deploy, publish, and observe Google ADK agents consistently.
• Docs as a product surface: MkDocs Material site with guides, references, and persona-aware navigation.

Quick start

uv sync --extra dev
uv run agents-cli --help
uv run agents-cli scaffold create my-test-agent --prototype
uv run agents-cli lint my-test-agent
uv run agents-cli eval run samples/sample_evalset.json

Parts

• CLI package: src/agents_cli
• Project templates: templates/
• Installable skills: skills/
• Documentation site: docs/ and mkdocs.yml

Command surface

agents-cli scaffold create|enhance|upgrade
agents-cli eval run|compare
agents-cli infra single-project|cicd
agents-cli deploy
agents-cli publish
agents-cli run
agents-cli playground
agents-cli lint
agents-cli install
agents-cli login
agents-cli info
agents-cli update

Skills package

The skills layer can be installed with:

npx skills add google/agents-cli

or locally with:

npm run skills:add

Included skills:

• google-agents-cli-workflow
• google-agents-cli-scaffold
• google-agents-cli-adk-code
• google-agents-cli-eval
• google-agents-cli-deploy
• google-agents-cli-publish
• google-agents-cli-observability

Verification

This repository has been exercised end to end:

uv sync --extra dev --extra docs
uv run agents-cli --help
uv run agents-cli scaffold create my-test-agent --prototype
uv run agents-cli lint my-test-agent
uv run agents-cli eval run samples/sample_evalset.json
uv run ruff check src templates
uv run mkdocs build --strict