Google Antigravity release notes: 2.3.0 queues messages and CLI 1.1.2 reviews new files

Antigravity received two updates on July 13 that fit together unusually well. Version 2.3.0 lets you queue the next message while the agent works. CLI 1.1.2 makes a new file easier to inspect in full screen and ensures scripts fail clearly when model selection or authentication does not work. The useful change is operational: these updates keep existing agent work in order.

Google Antigravity is Google's agentic development environment and CLI. An agentic development environment lets an AI agent read a project, use tools, and complete multistep jobs in the same workspace. A message queue is an ordered list of instructions waiting while the current job finishes.

Google Antigravity release notes: 2.3.0 queues the next instruction

Google Antigravity 2.3.0 is dated July 13, 2026. The changelog adds queued messages, settings for configuring how messages execute, and a Send Now option when a message needs to run immediately. The Queued Messages card also shows a count and sits in the chat input area. Google rolls out new versions gradually, so the update may take a few days to reach every user.

This is useful for anyone who currently waits for the agent or interrupts it with the next thought. You can let a build or analysis finish while queuing the follow-up. The release can also attach and render plain .txt files in conversations.

The stability work matters just as much. Antigravity retries intermittent backend overloads, stops background work when a conversation is archived, and fixes hangs after subagents finish. Diffs for files longer than 1,000 lines should also render correctly.

Human step: confirm that 2.3.0 has reached your installation. Queue the next instruction while a conversation is working and use Send Now only when it should run immediately. Google documents the feature and its settings but does not give an exact menu path in the changelog.

Source: Google Antigravity 2.3.0 changelog

CLI 1.1.2 makes new files and scripts easier to review

Antigravity CLI 1.1.2 was released on July 13, 2026, at 23:26 UTC. In the review screen for a new file, you can now press f to open a full-screen diff, matching the overview already available for edits to existing files. That detail helps when the agent creates a longer configuration file, script, or document from scratch.

Print mode, -p, can now receive the OAuth authorization code through the controlling terminal when standard input is already consumed by a piped prompt. Truly headless runs should fail fast with an actionable message instead of blocking. If --model points to a model that cannot be resolved, print mode exits with a non-zero code and lists available models. Interactive sessions keep the fallback behavior with a warning.

The CLI release also fixes MCP cleanup for custom agents, permission checks for nested command substitutions, and performance in conversations longer than 5,000 steps. MCP, or Model Context Protocol, is a standard way to connect an agent to external tools and data sources.

Human step: after updating to CLI 1.1.2, press f when the new-file review screen appears. Existing print-mode automation should capture the exit code and stderr so model or authentication failures reach the run log. The release notes verify f, -p, and --model; they do not specify an installation command.

Source: Antigravity CLI 1.1.2 release notes

What Nordic teams should test now

Choose one focused workflow where the agent already runs for a few minutes, such as a test run or an integration review. Queue a follow-up that only reviews the result and writes a short handoff receipt. In the CLI, use the full-screen diff for new files and confirm that automated runs report failures instead of getting stuck.

For Tool Forge work, connect the queue, review step, and run log, so the agent can keep moving without losing the next instruction or hiding a failure in the terminal.

Short example: use the new Antigravity feature

Queue this prompt while Antigravity is working on an active change. It contains no installation or interface steps.

When the current job finishes, review the diff and test output.
Return the changed files, any failures, and one recommended next action.
Do not start another change.

Good output should include:

• The files that actually changed
• Test failures or uncertainty that needs follow-up
• One recommended next action
• No new file changes

What to watch next

The changelog does not say where the queue execution settings live or how several waiting messages are prioritized. Those details are worth checking once 2.3.0 reaches more installations. For the CLI, the next practical check is whether the clearer failures work consistently in your CI jobs, remote terminals, and local scripts.

Sources: Google Antigravity 2.3.0 changelog, Antigravity CLI 1.1.2 release notes