A comprehensive, field-level mapping between Claude Code and Google's AI stack — Gemini CLI, Apps Script, ADK, and the broader Gemini ecosystem — for practitioners migrating AI-agent operations stacks.
This is not a feature comparison. It's a working reference that documents how every major system-level concept in Claude Code maps to its Google equivalent — including exact config field names, CLI flags, file paths, policy engine syntax, and behavioral gotchas that only surface in practice.
Last verified: 2026-03-17 Covers: Claude Code (Sonnet 4.x / Opus 4.x era) and Gemini CLI (current), Gemini 2.5/3.x models
You've built something on Claude Code — skills, automations, MCP integrations, subagent dispatch patterns — and now you need to transpose that system onto a client or team running Google's stack. This document makes that transposition efficient instead of painful.
Claude Code is monolithic. Google's equivalent is distributed. Claude Code is one tool — configuration, automation, multi-agent dispatch, MCP integration, and interactive use all happen in a single CLI agent. Google spreads this across Gemini CLI (closest 1:1 match), Apps Script (cloud automation), ADK (multi-agent orchestration), Cloud Functions (serverless scheduling), Vertex AI (enterprise deployment), Gems (custom instructions), and NotebookLM (knowledge synthesis).
The surprise finding: Gemini CLI is far closer to Claude Code than expected. It has GEMINI.md (≈ CLAUDE.md), SKILL.md files with the same frontmatter schema, MCP support with the same transport types, subagent definitions in markdown, headless mode (gemini -p), a policy engine, custom commands, hooks, and a memory system. The Gemini CLI sections are near-1:1 mapping exercises. The Apps Script, ADK, and Cloud sections cover the distributed layer that has no single Claude Code equivalent.
| # | Section | What it covers |
|---|---|---|
| 1 | Executive Summary | Paradigm overview, key behavioral differences |
| 2 | Architecture Overview | Component map, decision framework for which Google product to use |
| 3 | Configuration & Project Setup | CLAUDE.md → GEMINI.md, settings, MCP config, permissions, hooks, memory |
| 4 | Skills | SKILL.md format, discovery, activation, worked example |
| 5 | Automation & Scheduling | claude -p → gemini -p, three-tier automation framework |
| 6 | Multi-Agent / Subagent Patterns | Dispatch models, parallel execution, ADK comparison |
| 7 | MCP & External Tools | Transport types, per-connector migration path |
| 8 | Tool Calling | Read/Write/Edit/Glob/Grep → Gemini equivalents, API schema differences |
| 9 | Context Management | Window sizes (200K vs 1M+), compaction, caching |
| 10 | Non-Coding Use Cases | Knowledge management, business automation, Gems vs Projects |
| 11 | Apps Script Deep Dive | Triggers, quotas, Gemini API from Apps Script, state persistence |
| 12 | iOS/Mobile Integration | Gemini Live, Shortcuts → Apps Script bridge |
| 13 | Enterprise Considerations | Editions, admin controls, Code Assist, NotebookLM Enterprise |
| 14 | Migration Checklist | Step-by-step with phases, gotchas list |
| # | Appendix | What it covers |
|---|---|---|
| A | Tool Translation Reference | Parameter-level mapping for every Claude Code tool |
| B | Starter settings.json | Drop-in Gemini CLI config template |
| C | Automation Wrapper Script | Shell script template replacing claude -p with gemini -p |
| D | Skill Translation Example | Complete before/after of morning sweep command + comms agent |
| E | Apps Script Template | Runnable Apps Script project for cloud automation |
| F | Flag Cheat Sheet + Model Mapping | CLI flags, model equivalents, pricing comparison |
If you're in a hurry:
- Start with the Architecture Overview to understand which Google product maps to what
- Read the Migration Checklist for the full step-by-step
- Use the Flag Cheat Sheet to translate your automation scripts
This is a living document. These tools ship breaking changes weekly — corrections, updates, and additions are welcome. Open a PR or issue.
Built by Jim Prosser / Person Familiar