Google Antigravity's /swarm coordinates AI agents in parallel
A new slash command splits a prompt across a coordinator, isolated workers and adversarial verifiers, and the rollout widens paid access to Gemini 4 Carbon.
Google started rolling out /swarm in preview on 10 November, a new command in its Antigravity app and CLI that turns one prompt into a coordinated group of agents instead of a single one. A coordinator spins up specialised workers, each in its own isolated git worktree so their edits can’t collide, alongside adversarial verifiers whose job is to run tests and challenge the other agents’ diffs. Everyone talks over a shared, append-only message board, parking at zero token cost until another agent needs their input.
/swarm build a Three.js exhibition from these photos and engineering plates
The same rollout widens access to Gemini 4 Carbon, the model Google first limited to vetted security teams in October. It’s now callable through the paid tier of the Gemini API in Google AI Studio and Android Studio, and usable by Google AI Ultra subscribers in Antigravity, the Gemini app and AI Mode in Search. Carbon runs a 1-million-token context window and costs $2 per million input tokens and $10 per million output tokens, including thinking tokens, through 30 November, rising to $4 and $20 from 1 December.
Why it matters: multi-agent coding has mostly meant wiring up your own orchestration. A first-party command with worktree isolation and built-in adversarial review lowers the bar for trying it. If you’re evaluating Carbon on the API, the pricing doubles in under three weeks, worth running the numbers on before then rather than after.
The caveat: /swarm is preview-only and gated to Google AI Ultra subscribers, and Carbon isn’t on the free API tier at all.