It will be good if we can just run Antigravity on the cloud where your laptop is free for other tasks.
Also, the ability set up autonomous dark factory or orchestrator workflow on the cloud.
At least get user parity with Devin.
That would be the Antigravity SDK, which the Antigravity tools are using behind the scenes. And you can get some degree of autonomy with Gemini Spark actions. (I have spark send me daily summaries of social updates for example.)
Hi @sirinath,
Thanks for sharing this feedback! While Google Antigravity does not currently offer a turnkey, Google-hosted cloud sandbox where compute is provisioned for you automatically, we have logged your feature request for native cloud-hosted execution with the product team.
In the meantime, you can already achieve both cloud execution (keeping your laptop completely free) and autonomous multi-agent / orchestrator workflows today using the following capabilities:
You can run the Antigravity CLI (agy) on any Linux cloud VM, VPS, or container (e.g., Google Cloud Compute Engine, AWS EC2, DigitalOcean, or a Docker container) and control it remotely from your browser:
-
Install the Antigravity CLI (
agy) on your cloud Linux VM and sign in (agy). -
Start the background Remote Control daemon on the VM: (Works both with
systemdand in headless Linux containers withoutsystemd.)agy remote-control start --name my-cloud-vm -
Open https://antigravity.google.com/ from any desktop or mobile browser to launch conversations, monitor progress, and review diffs—even with your laptop turned off or closed.
- Docs: Remote Control | Google Antigravity Docs
- (Note: If you use the Antigravity desktop IDE, you can also connect to remote cloud VMs via Remote - SSH or Dev Containers so all builds, tests, and agent commands run on cloud compute.)
-
Set Up Autonomous “Dark Factory” & Multi-Agent Orchestrator Workflows
- Antigravity includes several built-in orchestration modes, headless CLI flags, and a Python SDK for long-running autonomous workflows:
/goal <task>(Autonomous Long-Running Execution): Instructs the agent to break a task into verifiable sub-goals and work autonomously until every requirement passes an automated completion audit. ( Slash commands overview | Google Antigravity Docs )/teamwork-preview <goal>(Multi-Agent Team Orchestration): Launches a coordinated team of specialized subagents—including an Orchestrator, parallel Explorers, an Armed Worker, Critics, Challengers, and a Victory Auditor—that plan and execute milestones with adversarial verification. ( Teamwork agent teams (/teamwork-preview) | Google Antigravity Docs )/boost <task>(Hierarchical Manager-Worker Orchestration): Runs a 3-tier manager-worker hierarchy that decomposes complex goals into dependency waves, dispatches isolated worker subagents in parallel, and verifies results. ( Boost deep reasoning (/boost) | Google Antigravity Docs )/schedule & /automation(Recurring Background Schedules): Set up recurring 5-field cron schedules or timers for continuous background tasks. ( Slash commands overview | Google Antigravity Docs )Headless CLI(agy -p) &Python SDK(google-antigravity):- Run non-interactive CI/CD or cloud daemon pipelines using agy -p “” --output-format json (or stream-json), which runs with no default timeout until the turn completes. ( Headless mode - Antigravity CLI | Google Antigravity Docs )
- Build custom programmatic multi-agent orchestrators in Python with pip install google-antigravity. (Antigravity SDK overview | Google Antigravity Docs)
- Antigravity includes several built-in orchestration modes, headless CLI flags, and a Python SDK for long-running autonomous workflows: