gws: The AI-Ready Google Workspace CLI

gws is a dynamic command-line interface for Google Workspace that automatically generates its command surface using Google's Discovery Service. It is specifically designed to support both human developers and AI agents through structured JSON outputs and integrated agent skills. The tool offers robust authentication options and advanced features like MCP server support, making it a powerful bridge for LLMs to interact with Workspace data.

Key Points

• Dynamic Command Generation: Uses Google's Discovery Service to build its interface at runtime, ensuring automatic compatibility with API updates.
• Dual-User Optimization: Provides a user-friendly experience for humans with help and completion, and a machine-readable experience for AI via JSON and skills.
• Flexible Authentication: Supports multiple workflows including OAuth, service accounts, and token-based access for local or CI/CD use.
• AI Integration: Features built-in support for MCP and Gemini CLI, allowing LLMs to interact directly with Workspace tools.
• Advanced Data Handling: Includes features like auto-pagination, multipart uploads, and response sanitization via Google Cloud Model Armor.

Sentiment

Mixed-to-cautiously-positive. The community is enthusiastic about the concept and the CLI-over-MCP pattern but deeply frustrated by the execution, particularly Google's painful authentication flow. The prevailing tone is 'this is exactly what we need, but it is practically unusable right now.' Trust in Google to maintain the project long-term is very low, tempered only by recognition that the underlying approach is sound.

In Agreement

• A CLI for Google Workspace with agent support fills a genuine need that the community has wanted for years
• CLIs are the right interface for AI agents — more ergonomic, composable, and lower token cost than MCP servers
• Dynamic command generation from Google's Discovery Service is a clever approach that keeps the tool automatically up-to-date with new API features
• The author's blog post on designing CLIs for AI agents provides valuable insights on the emerging pattern of agent-first tool design
• The AI agent wave is pushing companies toward building better APIs and documentation, which benefits human developers too

Opposed

• Authentication setup is a dealbreaker — requiring GCP project creation, OAuth app configuration, and verification is absurdly complex compared to simpler tools like GitHub CLI
• The 'not an officially supported Google product' disclaimer, combined with Google's history of abandoning projects, makes it risky to depend on
• Agent-first design makes the tool human-unfriendly — JSON params and lack of inline documentation hurt manual usability
• NPM distribution for a Rust binary is an odd choice that adds unnecessary dependencies and security concerns, with a 750MB install size
• Dynamic command generation means no static documentation exists, making it difficult for users to discover available commands and options
• The tool reportedly does not work with personal Gmail accounts, significantly limiting its audience