Google Antigravity 2.0 Splits Agent Work Across Surfaces
Google Cloud's June 10 developer guide frames Antigravity 2.0, Antigravity CLI, Antigravity IDE, and Antigravity SDK as different surfaces for agentic work. The message is that agent adoption now depends on choosing the right interface for the job.
Technical Signals
- Surface Choice: CLI, IDE, SDK, and app surfaces serve different automation and review patterns.
- Developer UX: Interactive coding, scripted workflows, and embedded product agents need different guardrails.
- Platform Pattern: Teams should standardize shared identity, logging, policy, and tool permissions across all surfaces.
- Rollout Risk: Without surface boundaries, agents drift into unmanaged scripts, editor plugins, and product features.
What Changed
Google's framing treats agentic development as a multi-surface product problem. A developer might want fast terminal automation, an IDE assistant for code context, an SDK for product integration, and a managed workspace for longer tasks. Each surface has a different failure mode.
Architecture Impact
The shared layer matters more than the shell. Identity, repository permissions, tool access, trace retention, prompt templates, and approval policy should be centralized so teams do not reinvent governance for every interface. The surface should change the experience, not the control model.
How To Choose
Use the CLI for repeatable local workflows, the IDE for contextual code changes, the SDK for embedding agents in internal platforms, and the full product surface for longer-running tasks with visible state. Teams should document these boundaries so developers know where each workflow belongs.
Governance Checklist
Inventory every agent surface in use. Require ownership, logs, model routing, permissions, secrets handling, and retention rules. Then measure which surface actually reduces cycle time without increasing review defects.