A First Look at Google Antigravity: The All-in-One AI Development Environment
Welcome to a groundbreaking new era of software development. Google DeepMind has unveiled Antigravity, a revolutionary product designed from the ground up for developers. This innovative platform seamlessly combines the best of all worlds into a single, cohesive surface: it’s a powerful code editor, an intelligent agent management system, and an integrated browser, all rolled into one. Let’s dive into what makes Antigravity a potential game-changer for how we build software.
Getting Started: A Smooth Onboarding
[00:23.012] [Welcome to Antigravity onboarding screen.]
The journey begins with a straightforward onboarding process. After launching the Antigravity app, you’re greeted with a welcome screen that guides you through the initial setup.
[00:27.794] [Theme selection screen in Antigravity.]
Personalization is key, and Antigravity lets you choose your preferred coding environment theme right away. Whether you’re a fan of dark mode or, like the presenter, prefer a light theme, you can select from options like Default Dark, Solarized Dark, Default Light Modern, and Solarized Light.
[00:32.747] [Google account sign-in page for Antigravity.]
Access is managed securely through your existing Google account. A simple “Sign in with Google” button authenticates your session, seamlessly linking the app and redirecting you back to the main interface once you’re successfully logged in.
The Three Pillars of Antigravity
[00:42.128] [Welcome to the Agent Manager introduction screen.]
Once logged in, Antigravity introduces its core concepts. The platform is built on three primary surfaces where you can get your work done: the Agent Manager, the Editor, and the Browser.
[00:52.288] [The Agent Manager inbox view in Antigravity.]
The first and central hub is the Agent Manager. This is your command center for creating, managing, and orchestrating AI agents across all your different workspaces. Think of it as a unified inbox where you can start new conversations (tasks) and monitor the progress of various agents working on long-running projects or deep research.
[01:06.182] [The integrated code editor in Antigravity.]
When you need to get hands-on with the code, the Editor is just a keystroke away. By pressing Command + E or clicking the “Open Editor” button, you transition into a full-featured coding environment. This isn’t just a simple text editor; it includes everything developers expect, such as tabs, intelligent auto-complete, and an agent sidebar, allowing you to seamlessly take over a task from an agent and bring it from 90% to 100% complete yourself.
[01:29.355] [The integrated browser testing a flight tracker application.]
The third pillar is a brand-new offering: an integrated Browser. This isn’t just any web browser; it’s a powerful tool with an AI agent built directly into its core. You can ask the agent to perform tasks like “test my feature,” and it will spawn a browser window, navigate to your app, and interact with it—clicking buttons, scrolling, and filling out forms—to test the very products you’re building. This brings the rich context of web applications and documentation directly into the agent’s workflow.
[01:46.402] [Agent-assisted development configuration screen.]
Before diving in, you can configure how you want to interact with Antigravity. The recommended mode is agent-assisted development.
In this mode, the agent uses its own judgment on when to run terminal commands locally and when to request a review for making changes. The LLM automatically decides if a task is simple enough to complete on its own or if it requires your attention for more complex decisions or questions about the codebase.
Building a Flight Tracker App from Scratch
[02:18.995] [Antigravity’s home screen with an empty inbox.]
To demonstrate Antigravity’s power, the video walks through building a flight tracker app. The process begins by creating a new local workspace, which is as simple as creating a new folder on your machine.
[02:41.050] [Initial prompt for building the flight lookup Next.js web app.]
With the workspace set up, the first step is giving the agent a prompt. The task is to build a Next.js web app where a user can look up a flight number and see its details.
Build me a flight lookup nextjs web app where the user can put in a flight number and the app gives you the start time, end time, time zones, start location, and end location of the flight. For now, use a mock API that returns a list of matching flights. Display the search results under the form input.
[03:03.022] [Terminal command execution mode options.]
Because the agent-assisted development (or “Auto”) mode was selected, the agent can run necessary terminal commands without requiring approval for every step. This autonomy allows it to efficiently set up the project structure. If a command were more sophisticated or potentially risky, the agent would pause and ask for user approval, building trust over time.
[03:22.388] [Terminal output showing the creation of a new Next.js app.]
The agent immediately gets to work, initializing a new Next.js application by running the npx create-next-app command with the appropriate configurations for TypeScript, Tailwind CSS, and ESLint.
Understanding Artifacts: The Agent’s Building Blocks
A core feature of Antigravity is the concept of artifacts. These are documents, typically in markdown format, that the agent creates to plan, track, and summarize its work. They are accessible in the right sidebar and provide incredible transparency into the agent’s process.
[03:50.855] [The agent’s “Task” artifact, a checklist of steps.]
There are three main types of artifacts:
- Task List: A simple checklist the agent maintains to keep track of its own progress.
- Implementation Plan: A detailed report generated before making code changes. It outlines the research, proposed changes, and the overall plan for the user to review and approve.
- Walkthrough: A summary document created after a task is complete. It details what was done, how it was verified (often with screenshots or recordings), and provides a final report.
[04:09.957] [The “Implementation Plan” artifact for the flight lookup feature.]
By reviewing the Implementation Plan, you can see a breakdown of the goal, the proposed changes (including new files and components), styling plans, and even a verification strategy. This gives you full control to approve the agent’s direction before it writes a single line of code.
Parallel Processing: Multitasking with AI
[06:24.380] [A new prompt to research the Aviation Stack API.]
One of the most powerful aspects of Antigravity is the ability to run multiple tasks in parallel. While one agent is building the main app, you can start new conversations to handle other tasks. For instance, you can ask another agent to research a live flight data API to replace the mock one.
Research the aviation stack flights API. I already have an API key if you want to test with curls. Lookup the documentation and make sure you use the curl responses to get the interface.
My key is ...
[06:50.146] [A prompt asking the agent to design app logos.]
Simultaneously, you can leverage Antigravity’s integration with Google DeepMind’s image generation models, like NanoBanana, to create a logo for the app.
Design a few different mockups for a logo for our app. I want one that's minimalist, one that's more classic, one that is clearly a calendar, and any others that you think might fit. I want to use this as the favicon for our app.
[07:09.612] [Four different app logos generated by the agent.]
The agent quickly generates several logo options based on the prompt’s requirements. After reviewing them, you can give feedback to select and implement your favorite one. This demonstrates how you can delegate both coding and creative tasks to run concurrently, dramatically speeding up your workflow.
Reviewing, Commenting, and Iterating
[08:14.651] [The research artifact for the Aviation Stack Flights API.]
Once the API research task is complete, the agent presents its findings in a new artifact. This research is incredibly thorough: it includes an overview, a test curl request, a sample JSON response, and even generates the necessary TypeScript interfaces based on that response.
[08:56.550] [Commenting directly on an artifact to provide feedback.]
Just like in a Google Doc, you can highlight parts of an artifact and leave comments. This collaborative process allows you to guide the agent’s next steps with precise instructions. For example, you can comment on the test request to remind it to use the API key you provided. After reviewing the plan and leaving feedback, you can approve it with a final comment.
[09:57.966] [The final generated code for the Aviation Stack utility.]
The agent then implements the utility, creating a new aviationStack.ts file complete with interfaces and a fetch function. At this point, you can jump into the Editor to integrate this new live data utility into the main application, replacing the old mock data logic. The editor’s AI-powered autocomplete, aware of the new file, makes this integration quick and error-free.
The Final Polish: Testing and Integration
[11:19.865] [The finished Flight Tracker application UI in the browser.]
With the live data flowing, the final step is to add a “killer feature”: integrating with Google Calendar. A new prompt instructs the agent to make each flight result card clickable, opening a pre-filled Google Calendar event.
For each flight result, make the entire card clickable to open a Google calendar link with the flight information (notably the times and locations). Test this with the browser and show me a walkthrough when you're done.
[12:20.201] [Agent-controlled browser testing the clickable flight card.]
The agent gets to work, modifying the UI components and then launching the integrated browser to test the changes automatically. You can watch in real-time as the agent’s cursor (indicated by a blue highlight) navigates the app, clicks a flight card, and verifies that it opens the correct Google Calendar link.
[12:43.350] [The final “Walkthrough” artifact showing successful Google Calendar integration.]
Finally, the agent produces a Walkthrough artifact summarizing the changes made, the verification steps it took, and even includes a screen recording of the test. This provides a comprehensive and verifiable record of the completed task.
[13:15.366] [Generating an automatic commit message in the source control panel.]
With all the features implemented, committing the code is the last step. Antigravity’s editor includes a source control panel with a “Generate commit message” button. Using the context of the conversation and the file changes, it generates a perfect, detailed commit message for you.
Conclusion
In just a few minutes, a complete, full-stack application was built from scratch, showcasing a new paradigm for development. This workflow—centered on reviewing artifacts, managing multiple parallel agents, and finishing tasks with an AI-powered editor—represents the future of getting work done.
The Antigravity team is already working on what’s next, so stay tuned for more exciting updates. This is just the beginning of a whole new way to build.