Google CLI Links OpenClaw to Gmail Unsupported

Google Workspace CLI, an open-source package the company posted on GitHub in March 2026, offers a first-party path to connect generative AI agents—including OpenClaw—to Gmail, Drive, Calendar, Docs, Sheets, Slides, Google Chat, and other Workspace surfaces. Repository language reviewed by Ars Technica states the project is not an officially supported Google product, and Google warns functionality may shift as underlying APIs change.

For developers, the draw is structured JSON output, a large catalog of pre-built agent skills, and OAuth inside Google's flows instead of ad hoc screen-scraping. For security and compliance teams, the same power means any misconfigured or over-permissioned agent can still mis-handle mail, attachments, or calendars whether the connector is first-party or not.

What the CLI bundles

Ars Technica's March 2026 walkthrough places the project under Google's GitHub organization, highlights JSON suited to machine parsing, and notes Google Cloud director Addy Osmani said in a social post the tree already included more than 40 skills at launch. The same reporting describes an MCP server mode for assistants such as Claude or Google's Gemini command-line tooling for teams that want that stack.

OpenClaw stays separate; Google documents the hook

OpenClaw, the agentic automation layer that surged in early 2026, remains its own project. Google's addition is documentation and integration paths so OpenClaw can drive Workspace actions through the new CLI rather than fragile one-off scripts.

That distinction matters for procurement: enterprises can see which APIs are touched, how OAuth scopes are granted, and where support boundaries sit—even when the product tier is explicitly unsupported.

Unsupported status and operational risk

Ars quotes repository language that Workspace CLI is not backed by enterprise support the way paid Workspace features are. Early adopters accept that automations built in pilots can break when Google changes an endpoint.

Industry coverage in 2026 has repeatedly flagged prompt-injection and over-permissioned agents; pairing OpenClaw-style autonomy with live inboxes raises the same class of issues the CLI does not remove.

Before production

Leads should still map least-privilege OAuth scopes, log automated writes, sandbox agents before they touch regulated data, and require human confirmation for irreversible sends or deletes. Watch whether Google promotes the project to supported status and how Workspace administrators set allowlists for agent tokens tied to the tool.