Route multiple backing models through Antigravity CLI.

Antigravity routes Gemini, Claude, and GPT-OSS backing models through one agent. Model labels choose the backing model while ProPR owns planning, isolated execution, and the GitHub PR loop.

Agent route ready
Approved planContext and scope attached
Gemini / Claude / GPT-OSS Isolated worker
GitHub pull requestCommits, logs, and follow-ups
Your credentials Task history kept GitHub merge decision
Choose the agent per phase. The plan, branch, and pull request keep the work legible.

Agent fit and access

Best fit

Fast implementation, context-heavy planning, and cost-sensitive review or analysis tasks.

Access

Authenticate Antigravity on the host and reuse that access inside agent containers.

Role in ProPR

Antigravity CLI participates in the same plan, task, PR, follow-up, logging, and recovery model as other agents.

Three vendors through one agent.

All Antigravity models run with a 1M-token context window through the same subscription and credential mount. Use the exact identifiers below in model configuration; managed issue labels provide GitHub routing.

ModelModel identifiersReach for it when
Gemini 3.7 Flash (High / Medium / Low)antigravity-gemini-3.7-flash-high
antigravity-gemini-3.7-flash-medium
antigravity-gemini-3.7-flash-low
Fast, cost-efficient passes — the tier trades speed against reasoning effort.
Gemini 3.1 Pro (Low / High)antigravity-gemini-3.1-pro-low
antigravity-gemini-3.1-pro-high
Context-heavy planning, analysis, and implementation.
Claude Sonnet 4.6 (Thinking)antigravity-claude-sonnet-4.6-thinkingCareful reasoning without leaving the Antigravity subscription.
Claude Opus 4.6 (Thinking)antigravity-claude-opus-4.6-thinkingThe hardest changes routed through the same access.
GPT-OSS 120B (Medium)antigravity-gpt-oss-120b-mediumAn open-weight option for standard passes.

Bring your own Google AI plan — ProPR never marks up tokens.

Antigravity authenticates with your Google account (agy login). Regular use needs a Google AI subscription, with pay-as-you-go credits for extra capacity. You pay Google directly; ProPR adds no markup.

PlanPriceGood for
Google AI Pro$20 / moEntry tier for regular Antigravity use.
Google AI Ultra$100–$200 / mo5× to 20× higher Antigravity usage limits.
Pay-as-you-go creditsMeteredTop-up credits for Antigravity beyond plan limits.
Compare Google AI plans and subscribe →
  • Model choice

    One Antigravity entry exposes Gemini, Claude, and GPT-OSS backing models — see the table above — usable through managed issue labels, plan settings, and review controls.
  • Phase routing

    Use Antigravity CLI for planning, implementation, review, or follow-up changes.
  • Access model

    Use the Antigravity credentials configured in your self-hosted environment.
  • Usage tracking

    Enable Agent Tank and ProPR shows Antigravity CLI usage around tracked calls.
The ProPR agent playground model picker with Claude, Codex, Mistral, Antigravity, and OpenCode model chips selectable for side-by-side tests.
The playground exposes Antigravity backing models beside Claude, Codex, Mistral, and OpenCode.

Setup details live in the docs.

ProPR runs the official Antigravity CLI, and the setup specifics — CLI versions, model aliases, and credentials — are covered in the documentation. Inside ProPR, Antigravity CLI follows the same path as every other agent: planning, isolated implementation, pull request, follow-up, and recovery.