Google Antigravity adds embedded terminals, Git tools, and generative UI

Google Antigravity shipped a concentrated set of late-August updates across its desktop product, CLI, and Python SDK. Version 2.10.0 adds embedded development tools; version 2.11.0 adds generative UI; and adjacent CLI and SDK releases improve remote control, voice, model selection, and custom-tool workflows.

Embedded terminal and Git-native version control

Antigravity 2.10.0, dated August 24, introduces an embedded terminal and Git-native version control directly in the sidebar. The same release adds audio file attachments, interactive image commenting, rich MCP tool-execution previews, and performance improvements.

Keeping these tools next to the agent workspace can reduce context switching, but teams should still review the commands, repository state, and MCP actions before allowing an agent to make consequential changes.

Inline interactive artifacts

Antigravity 2.11.0, dated August 26, adds generative UI that renders HTML artifacts inline in chat. Google’s accompanying product post describes interactive HTML/CSS/JavaScript components, live charts, diagrams, and visual representations of complex data that can be inspected in the artifact preview and exported as standalone HTML.

The visual tooling is intended to let an agent prototype and iterate on an interactive result instead of stopping at markdown or a static image. The product post says the visualizations render on the user’s machine without downloading additional packages or dependencies and can run offline.

CLI and SDK improvements

Late-August CLI releases add /voice dictation, embedded ripgrep code search, status-line cost metrics, and UTF-8 fixes. The CLI also gains model-selection command arguments, reasoning-effort autocomplete, transient retry handling, dynamic remote-control port allocation, and more predictable headless behavior.

The Python SDK adds subagent-scoped custom tools, an on_compaction lifecycle hook, universal JSON Schema normalization for custom tools targeting local OpenAI-compatible endpoints, and Vertex authentication overrides. These features are useful for teams building custom agent workflows, but the changelog notes that new versions roll out gradually and may take a few days to reach all users.

September follow-up: Gemini 3.8 Flash and CLI updates

Antigravity 2.12.2, dated September 3, adds access to Gemini 3.8 Flash through ADC for AGY Enterprise users. Version 2.12.0, dated September 2, adds quoting, /boost, and several settings and chat improvements.

The Antigravity CLI also moved through versions 1.1.23–1.1.26 in early September. The official CLI changelog lists improvements to model selection and the /resume picker, Gemini 3.8 Flash in the model catalog for GEMINI_API_KEY users, MCP configuration parsing, headless stream cleanup, artifact navigation, and worktree/session reliability.

These releases extend the August desktop updates rather than replacing them. Because Antigravity rolls out versions gradually, teams should check the installed desktop and CLI versions before assuming every control is available.

Sources

• Google Antigravity — “Changelog” (versions 2.10.0–2.12.2 and CLI releases): https://antigravity.google/changelog
• Google Antigravity — official CLI changelog: https://github.com/google-antigravity/antigravity-cli/blob/main/CHANGELOG.md
• Google Antigravity — “Visualizing with the help of Antigravity” (August 26, 2026): https://antigravity.google/blog/visualizing-with-the-help-of-antigravity
• Google Antigravity on X — official product updates: https://x.com/antigravity