Google Antigravity release notes: codelab shows build workflows

Google still has not posted a new Antigravity version after 2.0.11 in the official changelog. The useful signal today is a Google Codelabs update instead: "Building with Google Antigravity" was updated on June 11, 2026, and shows how Antigravity can be used for web research, app builds, artifacts, and test work in the same agent flow.

Google Antigravity is Google's agentic development platform. In this context, agentic means that an AI agent receives an instruction, asks for permissions when needed, creates artifacts, and hands the result back for human review.

Source: Google Antigravity Changelog still shows 2.0.11 from June 3, 2026, as the latest version. Building with Google Antigravity is marked last-updated="2026-06-11T03:15:05Z" in Google's codelab page.

Google Antigravity release notes: what changed in practice

This is not a new 2.0.12 release. It is a documentation signal with release-note value: Google now describes Antigravity as a place where you can move from simple web research to full-stack apps and more advanced unit tests. The codelab is short, but the workflow matters more than the demo apps.

• Antigravity can let the agent visit a website, read information, and summarize it after the user grants access.
• The agent can create an implementation plan as an artifact before work continues.
• The user can review the plan, comment on it, and only then let the agent continue with Proceed.
• Once code has been created, the user can open it in Antigravity IDE through Open IDE.
• Google also shows a test case where the agent generates mocks, unit tests, and validates tests against an existing Python file.

Source: Building with Google Antigravity describes use cases for News Highlights, Python + Flask, a Pomodoro app, and "Generate Unit Tests, Mock Stubs and Validate Tests". The same source says the agent asks for permissions when needed, produces artifacts, and notifies the user when the task is done.

Why Nordic teams should care

For Nordic teams testing AI in everyday work, the point is practical: Antigravity is starting to look like a controlled workbench, not just a chat window next to the code. An agentic IDE is a development environment where the agent's plan, files, runs, and review points sit close to the same project. That makes it easier to see what the agent wants to do before anyone clicks ahead.

This fits small automation teams that already have recurring research, documentation, or testing tasks. Start with a workspace where the data is bounded. Give the agent read access first. If the flow later connects to real systems, use scoped API keys, environment variables, secret managers, approval gates, and audit logs instead of pasting sensitive values into chat.

Human step: do this before the prompt

Create an Antigravity project and add the right local folder. Google's codelab shows the pattern: create a folder, start a new project, add the folder, and start a conversation in the workspace. If the agent creates an implementation plan, review it before pressing Proceed. If code is created and you want to inspect the files, use Open IDE.

Source: Building with Google Antigravity shows project folders such as news-highlights and conference-website, and describes implementation_plan.md, the Proceed button, and Open IDE. Getting Started with Google Antigravity describes projects as a combination of folders that define the agent's environment and scope.

Short example: use the new Antigravity feature

Once the project exists and you have an empty workspace, give the agent a short task that tests the same pattern Google shows: browser access, artifact, permission, and human review.

Visit https://docs.cloud.google.com/release-notes and summarize the latest relevant Cloud Run and BigQuery changes for our automation backlog. Create an implementation plan with one workflow idea, required permissions, approval points, and how a human should verify the result before any deployment.

Good output should include:

• specific product names, dates, and source links from the release notes page
• a clear implementation plan, not just a summary
• which permissions or tools the agent needs before it does more
• a human review point before anything connects to production

Source: Google shows a related prompt pattern in the same codelab: Visit https://docs.cloud.google.com/release-notes and get me a summary of the release notes, categorized by product.

What to watch next

Watch the official Antigravity changelog for the next version entry after 2.0.11. If the next update touches the CLI, SDK, MCP servers, or clearer project permissions, the question is not only "what is new?" but "which part of the agent's work can we now make traceable?" That is where Tool Forge becomes relevant: small agent flows with the right access, clear stop points, and receipts people can review afterwards.

Source: Google Antigravity Releases says Antigravity and Antigravity IDE auto-update to the latest version by default, while staying on older versions requires manual or disabled update mode in settings.