How Do You Run Coding Agents in Parallel With Google Antigravity?

Hey hey hey, welcome to Agentic Pod! I'm Liam.

And I'm Stella! Liam, Google's Antigravity 2.0 runs several coding agents at once, and it has a free tier.

Yep. One curl command installs the CLI. Then you can have agents working while you get coffee.

Okay, I'm in. But, um, is the free part real? Or is it free like a free sample at the mall?

We'll get to the quota. First, what is it?

So is this an editor? Another VS Code thing?

Not anymore. Google showed Antigravity 2.0 at I/O on May 19. It's a standalone desktop app for running and managing agents, and it works without an IDE.

And the old editor?

It lives on as a separate Antigravity IDE. And there's a CLI, which replaced the Gemini CLI for consumers.

Wait, so the thing I installed last year is gone?

For consumers, yeah, the Antigravity CLI took its place. Current versions are 2.23.1 for the app and 1.3.6 for the CLI.

I just got used to the old name.

Everyone did. Now the fun part: how you try it.

Okay, give me the steps. I'll be the person who skips the docs.

On Mac, Linux or a Googlebook, you open a terminal and paste one curl line from the download page. It pipes the installer into bash.

And Windows people?

PowerShell has its own one-liner, same page. The desktop app needs macOS twelve or newer, Windows ten, or a recent Linux, on x64 or ARM.

Cool. Then what? I just ask it for stuff?

Open it inside a repo, and start with a small job. Say, "add tests for the date parsing module."

And the parallel bit?

You start a second agent on a different task, say a README fix. Each one runs on its own, and you watch them in the app.

Huh. So I'm basically a manager now.

A manager with zero meetings. That's the pitch.

Okay, so, what can go wrong? Because I feel like something can.

First, permissions. Turbo mode runs terminal commands without asking you.

Oh no. That sounds like me with a credit card.

Right. Strict mode asks for approval and runs commands in a sandbox with limited network access. Start in Strict.

Got it. And the second one?

Quota. Free users get a usage quota that refreshes weekly. Paid Google AI Pro and Ultra plans get bigger quotas that refresh every five hours.

So parallel agents eat the free quota faster.

Exactly, every agent you start spends from the same quota. So start small.

Is there something that shows me what an agent actually did?

Yes, and that's my favorite part. Agents produce Artifacts, like a task list, an implementation plan, screenshots, even browser recordings.

Okay, so should I switch?

If you've never run parallel agents, this is a nice free way to learn. The app, the CLI and the Artifacts make it easy to see what each agent did.

And if I'm happy where I am?

Then try one small task, in Strict mode, and compare. Don't move your whole workflow on day one.

And who's it not for?

Anyone who needs the IDE inside a company, because enterprise accounts can use the app and the CLI, but not the IDE.

Fair. That's actually a useful detail.

Quick recap. Install the CLI with one command. Start in Strict mode. Watch the quota. Read the Artifacts. Links to the guide and sources are in the description. Which task would you hand to a second agent first? Tell us in the comments. See you next time!