Google Antigravity 2.0 - Orchestrate multi-agent workflows from a desktop app

Google just separated the agent manager from the IDE and shipped it as its own desktop app.

What it is: Antigravity 2.0 is a standalone desktop application built entirely around orchestrating multiple AI agents in parallel, scheduling background tasks, and managing subagent workflows across projects.

Most AI coding tools still make you sit in the loop: prompt, wait, respond, repeat. Antigravity 2.0 breaks that pattern by letting agents run in the background on cron-like schedules, work in parallel across subagents, and carry full project context from AI Studio to your local environment in one click.

• Run multiple agents simultaneously across parallelized subagent workflows

• Schedule tasks that trigger agents automatically in the background

• Export full projects from Google AI Studio to local development with one click

• Connect natively with Firebase and Android

• Issue voice commands instead of typing prompts

• Use the CLI for terminal-native work or the SDK to deploy custom agents on your own infrastructure

If you're a software developer or engineering team that has outgrown one-shot prompting and wants agents running across your build loop without babysitting them, this is built for that workflow.

P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends

Never used the IDE version much either. The subagent workflow is where it gets interesting: running parallel agents without babysitting each one is the shift that makes this feel different from Cursor or Claude Code. The background scheduling is a nice touch too. Excited to see how far they push the Firebase and Android integrations.

how to make this bunch of ai agents run in the background so as not to allow burn all tokens my tokens at all?

Parallel agents are powerful, but the monitoring layer is what makes this actually usable. Once multiple agents work at the same time, the hard part becomes catching conflicts, knowing what changed, and deciding what needs human review.

Scheduled agent work gets interesting the moment a team comes back later and has to decide whether a run is safe to accept or needs review. I would want every scheduled task to leave a small receipt: owner, allowed capability set, files or services touched, stop reason, and whether it completed, paused, or hit a guardrail. Without that, background autonomy can create more context debt than leverage.

The "sit in the loop: prompt, wait, respond, repeat" problem is real. I've been running Claude Code for side projects and the context switching cost is higher than people admit — you context-load the codebase, fire off a task, then either stare at it or switch to something else and lose the mental model entirely.

The cron-like scheduling piece is what I want to understand better. If I schedule a refactor agent to run at 2am, what happens when it hits a decision that needs human judgment — does it pause and queue, or does it guess and commit? The difference between "runs in background" and "runs unattended" is huge for trust.

Also curious about the subagent orchestration — when one agent spawns three parallel workers, how does the parent agent reconcile conflicting outputs? Is it another LLM pass, or is there a deterministic merge strategy?

Google announced so many Antigravity updates lately SDK, CLI, etc.