Google Antigravity release notes: CLI 1.0.15-1.0.16 makes agent work more visible

Google Antigravity release notes can sound dry until you run several agents at once. Then the question gets simple: can you see what is running, find the right log, and stop the wrong run before it keeps going? CLI 1.0.15 and 1.0.16 are about that kind of control.

Google Antigravity is Google's agentic development platform, a workspace where AI agents can be launched, monitored, and coordinated around code and other work. Antigravity CLI is the terminal surface for that agent work. For Hammer readers, the useful news is not a bigger model. It is better visibility while agents are already moving.

Source: Getting Started with Google Antigravity

Google Antigravity release notes: what 1.0.15 and 1.0.16 change

Google published Antigravity CLI 1.0.15 on July 1 and 1.0.16 on July 2, 2026. Together, they point to a practical improvement: background jobs, subagents, and permissions are easier to read while work is in progress.

1.0.15 adds an interactive status indicator below the input box that shows active subagents and background tasks in real time. It also improves /permissions so the panel reloads configurations from disk and reduces the chance of overwriting changes by mistake. The MCP connection timeout increases to 60 seconds, which matters when a local or company MCP server starts slowly. MCP, Model Context Protocol, is a way to connect an AI agent to external tools and data sources through a more standard interface.

1.0.16 continues the same theme. The /tasks detail panel now scrolls to the bottom as new background task logs stream in and opens at the latest output, unless the user has manually scrolled up. The release also adds client-side retries for transient model errors, fixes dynamically defined subagents by moving definitions from JSON to Markdown, and removes crashes when background tasks or terminal commands return empty output, such as sleep.

Source: Antigravity CLI 1.0.15 release notes and Antigravity CLI 1.0.16 release notes

Why this matters for practical agent workflows

A normal chat box is fine for a single question. Once subagents, background tasks, MCP tools, and permission rules enter the workflow, the team needs a run receipt. What started? Which task is waiting? Which configuration is active? Which log should someone review before the next step?

That is where these Google Antigravity changelog items are useful. The status indicator and /tasks improvement make parallel work easier to follow. The /permissions change fits teams that manage permissions through files and still need the CLI to reload them cleanly. The longer MCP timeout makes integrations less brittle when a server needs a little more startup time.

For Nordic teams building automation, this moves Antigravity CLI closer to an operating workspace. You can let an agent work for longer, but still require clear stop points, scoped API keys, environment variables or secret managers for sensitive values, output redaction where needed, and human approval before the agent changes permissions or runs risky commands.

What to test today

Human step: open an updated Antigravity CLI project where you already have a real agent workflow. If you use MCP servers or project-level permissions, make sure the test is in the right repo and uses the right configuration files. Do not run a production-critical workflow just to test a release note.

Then run a focused review:

• Are active subagents or background tasks easier to spot through the status area?
• Does the /tasks view give you a useful log position as new output arrives?
• Does /permissions reload changes without overwriting them?
• Does the MCP connection need a clearer startup check now that the timeout is longer?

Short example: use the new Antigravity feature

Paste this once the project is already open in Antigravity CLI:

Review this Antigravity CLI project as an operations lead.
Focus on active subagents, background tasks, MCP startup reliability, and permissions that may change from disk.
Do not edit files or permissions yet.
Return a short run receipt: what is active, what needs human approval, and one change that would make the next agent run easier to review.

A good answer should show:

• which tasks, subagents, or logs the agent actually checked
• whether MCP or permissions need a clearer control point
• which parts can continue and which need human approval
• one concrete improvement to the next run receipt

The Hammer angle

This is a useful release to watch if you want to move from one-off prompts to agent workflows that can run in daily work. In a Tool Forge process, we would start with a run receipt, a permission map, and an MCP startup check before connecting more systems. Then Antigravity is not just a stronger terminal agent. It becomes part of a workflow people can review.