Google Workspace CLI opens Workspace to agents

The real story here is not another CLI, but a blueprint for how developer tools need to evolve when AI agents become first-class users.

• –Dynamic schema introspection means agents can discover methods, params, scopes, and payload shapes at runtime instead of relying on stale docs or prompt stuffing.
• –Structured JSON output, raw payload support, and an MCP surface make this far more usable for coding agents than ad hoc shell scripts or browser automation.
• –The project bakes in unusually strong agent safety rails, including `--dry-run`, field masks for context control, input hardening, and optional Model Armor response sanitization.
• –It also looks like a serious ecosystem play: the repo ships 100+ skills, appears in the Gemini CLI extensions catalog, and is explicitly designed for multiple agent surfaces.
• –The main caveat is maturity: the repo is still marching toward v1.0 and is labeled as not officially supported by Google, so early adopters should expect some setup and breaking-change risk.