Best fit
Fast implementation, context-heavy planning and cost-sensitive review.
Gemini, Claude and GPT-OSS models. One Google login.
Antigravity CLI puts three vendors' models behind one agent. Pick the model per issue; ProPR handles planning, isolated execution and the GitHub PR loop.
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
Agent fit and access
Antigravity CLI plans and models
Antigravity CLI signs in with a Google account and runs Gemini 3.8 Flash at no cost under basic weekly rate limits. Google AI Pro at $19.99 a month unlocks Gemini 4 Argon, Claude Sonnet 5.5 and Claude Opus 5.5; Ultra at $99.99 or $199.99 multiplies the Pro limits by five or twenty and adds Gemini 4 Carbon. Google publishes five-hour and weekly windows and the multipliers, and no request counts.
| Plan | Price (USD) | Models | Published limits |
|---|---|---|---|
| Free | $0 | Gemini 3.8 Flash, GPT-OSS 120B (Google removes GPT-OSS on 18 November 2026). | Basic weekly rate limits. No number published. |
| Google AI Plus | $4.99 / month | Same models as Free. | Limited agent requests. No number published. |
| Google AI Pro | $19.99 / month | Adds Gemini 4 Argon, Claude Sonnet 5.5 and Claude Opus 5.5 (paid, non-trial subscriptions). | Five-hour and weekly windows. No number published; AI credits top up past the quota. |
| Google AI Ultra | $99.99 / $199.99 / month | Adds Gemini 4 Carbon. | 5× or 20× the Pro limits. |
| Business via Google Cloud | From $30 / seat / month, or pay-as-you-go | Enterprise model list, Gemini 3.1 Pro included. | Spend caps across API and subscription plans. |
Every plan runs headless: agy -p runs one task on the cached sign-in and exits, which is how ProPR drives it.
Sources: Antigravity models and plans, Google AI plans, Antigravity CLI headless docs. Verified November 2026.
That plan runs in ProPR on GitHub issues and pull requests. Label an issue llm-antigravity-argon or llm-antigravity-opus55 and ProPR plans the task, runs Antigravity CLI on your Google login in an isolated workspace, and opens the PR. ProPR lists the Gemini, Claude and GPT-OSS entries in its model picker; you pay Google directly and ProPR adds no token markup. See the guide to running a coding agent on GitHub issues, which covers Antigravity CLI alongside Claude Code.
Access
One Antigravity login on the host covers every run.
Role in ProPR
Antigravity CLI runs every phase, goals included, with Gemini, Claude and GPT-OSS models behind one Google login.
Three vendors through one agent.
Every Antigravity model gets a 1M-token context window on the same subscription. Pick the model, then the reasoning level: globally, per model, or per issue with a level-* label. Gemini 3.8 Flash is the default.
| Model | Issue label | Context | Notes |
|---|---|---|---|
| Gemini 3.8 Flash (default) | llm-antigravity-flash38 | 1M | Fast, cost-efficient passes; raise the reasoning level (low / medium / high) when the change needs more thought. |
| Gemini 4 Argon | llm-antigravity-argon | 1M | Planning and PR review; ProPR recommends it for both. Low / medium / high reasoning. |
| Claude Sonnet 5.5 | llm-antigravity-sonnet55 | 1M | Careful reasoning on the Antigravity login. Low / medium / high reasoning. |
| Claude Opus 5.5 | llm-antigravity-opus55 | 1M | The hardest changes, routed through the same access. Low / medium / high reasoning. |
| GPT-OSS 120B | llm-antigravity-gpt-oss-120b | 1M | An open-weight option for standard passes at medium reasoning. Google retires it from Antigravity on 18 November 2026. |
Gemini 3.1 Pro is retired: llm-antigravity-pro and saved Pro selections run on Gemini 4 Argon.
Model choice
One Antigravity entry covers the Gemini, Claude and GPT-OSS models above, selectable from issue labels, plan settings and review controls.Phase routing
Use Antigravity CLI for planning, implementation, review or follow-up changes.Goals
Antigravity runs Goals as a resumable session that takes your corrections between steps.Usage tracking
With Agent Tank, ProPR shows the Antigravity capacity each call spent.
Set up Antigravity CLI
Antigravity CLI setup, models and credentials in the ProPR docs → ProPR runs the official Antigravity CLI.