01What Is Antigravity Agent Teams?

Antigravity Agent Teams is a new update to Google's Antigravity AI coding agent that lets you create a whole team of specialized sub-agents from one command. Instead of working with a single AI worker that does everything step-by-step, you get a coordinator, multiple builders, and a verifier that work in parallel to plan, build, and check complex systems.

In this episode, Julian Goldie shows how this works by typing /teamwork-preview in the Antigravity CLI. The system then drafts a team sheet—like an org chart for your AI team—and once you approve it, the team starts working in the background. You can watch the whole process in the dashboard, and the plan is saved in your workspace so you can come back to it later.

This is something Julian calls the "one command company" because you can spin up a full team of AI specialists from a single prompt. The team plans, builds, and verifies the work, handing you a finished product without you having to manage every step.

02Key Features of Agent Teams in Antigravity

The main feature is the ability to create a dynamic team of specialized sub-agents that coordinate in the background. You get a coordinator that reads your task and drafts a team sheet, which lists the specialists to hire, what each one owns, and how they hand off work to each other.

Here are the key features shown in the episode:

  • Coordinator agent that drafts the team and reads the work plan.
  • Builder agents that work in parallel on different parts of the project.
  • Verifier agent that checks the work before delivery.
  • Team sheet that outlines roles and handoffs—you approve it once, and the team runs.
  • Dashboard view where you can see the live multi-agent collaboration, switch between sub-agents, and use template presets.
  • Workspace saving—the whole plan is saved locally so you can come back to it.

The update also lets you use custom APIs or even local models like Gemma 4 from Google, so you're not locked into one provider.

03How to Set Up Antigravity Agent Teams

Setting up Agent Teams is simple, and you don't need to be technical. Julian shows the exact steps in the episode. Here's how to get started:

  1. Make sure you have the Antigravity CLI set up. If you don't, you can get a system that has it already plugged in, like Julian's Agent OS.
  2. Run the update command in your terminal: agy update (that's short for Antigravity update).
  3. Open your Antigravity CLI and type /teamwork-preview to start a new team task.
  4. Describe your project. For example, "Create a team for a new blog about Hermes agent and AI automation."
  5. The system will draft a team sheet with specialists for your task.
  6. Approve the plan, and the team will start building in parallel. You can watch it live in the dashboard.

That's it. You don't need to write role prompts or split tasks by hand—the coordinator handles all of that for you.

04Using Agent Teams: Real Examples from the Episode

Julian demonstrates how to use Agent Teams with a few examples. In one test, he ran /teamwork-preview for an SEO blog, and the system quickly created a markdown file with a full plan for the agents. He then ran it live and spun up the agentic sprint directly in the dashboard.

In another example, he asked the team to create a page inside his Agent OS. The team sheet included an SEO architect, a Next.js developer, SEO Ops, and a QA developer (called Lighthouse X). The coordinator planned out exactly how the page would be built, and the multi-agent preview showed the orchestrator, sub-agents, and template presets.

You can also use a "team splitter" feature to select the project type, describe it, and the system assembles the team for you. The files end up saved locally, so you can open the dashboard and continue working later. Julian also shows how the agents can build playable games and tools right inside the guide.

05Results and Comparisons: Old Way vs. Agent Teams

Julian compares the old way of working with Antigravity to the new Agent Teams approach. Before, you had a single worker that did everything step-by-step, which took more time because there was no parallel processing. You only had quality control if you explicitly asked for it, and you had to split tasks and write role prompts by hand.

With Agent Teams, you get a coordinator that drafts the right specialists, builders that work in parallel, and a verifier that checks everything before delivery. Roles, briefs, and team handoffs are all written by the system, so you don't have to manage any of that.

Julian also addresses common concerns: some people say you need to be technical, but you just type one command. Some say more agents cost more, but the CLI is free to use. And some worry it's just a preview, but Julian argues the people who start building now will get ahead.

06Tips, Limits, and Mistakes to Avoid

Here are some tips and warnings from the episode to help you get the most out of Agent Teams:

  • Don't wait for the full release. Julian says people who start using the preview now will get ahead, and those who wait will fall behind.
  • Use a verifier agent. Without a quality control sub-agent, the output can get super messy. With one, the system focuses on making quality the best it can be.
  • You can plug in custom APIs or use local models like Gemma 4 to avoid extra costs.
  • Save your plan. The whole plan is saved in your workspace, so you can come back to it later—don't lose it.
  • Run in the background. You can have teams working while you do other tasks, so use that to your advantage.

One limit Julian mentions is that the feature is still in preview, so it may not be as polished as the final version. But he recommends starting now to get used to it.

07Who Should Use Antigravity Agent Teams and Why It Matters

Agent Teams is for anyone who builds apps, websites, games, or tools with AI. It's especially useful for people who want to move faster without having to manage every step of the process. You don't need to be technical—the command is simple and the system handles the project management.

This matters because it changes the way you work with AI. Instead of one AI agent doing everything step-by-step, you get a whole team that works in parallel. This speeds up production and adds a layer of quality control that wasn't there before.

Julian also points out that this is part of a bigger system he calls the Agent OS, which integrates Antigravity with other agents like Hermes, Claude, Codex, and GPT-5. If you want a more customized experience, you can get that system through his community, the AI Profitable Boardroom. But even on its own, Antigravity Agent Teams is a powerful update.