Google Antigravity

About

Key Features

• Describe a software task → agents plan steps and execute changes across your project
• Point it at a web app URL → built-in browser automation clicks through flows and captures screenshots and logs
• Choose Gemini 3.1 Pro/3 Flash, Claude Sonnet or Opus 4.6, or gpt-oss-120b → swap models per task without changing your setup
• Type as you code → get unlimited Tab completions and unlimited Command requests within weekly limits
• Run multiple tasks in parallel → asynchronous agents persist artifacts and task-level context for later review

Use Cases

• A full‑stack developer bootstrapping a new feature and letting an agent fix build errors after running the test suite
• A QA engineer validating a checkout flow end‑to‑end and exporting the agent’s plan, steps, and screenshots as test evidence
• A release manager kicking off nightly regression tasks and reviewing execution trails to spot flaky UI failures

Try It Like This

1. 1
   Bootstrap a new feature across repo

   Install the Antigravity desktop app (Mac/Windows/Linux) → point the agent at your repo and describe the feature in plain English; the agent plans tasks, creates branches, and submits code diffs for review → review diffs, request small edits, and merge once tests pass and you’ve inspected execution trails.

2. 2
   Fix a CI/build break automatically

   Open Antigravity and run the test suite so the agent sees failing logs and stack traces → ask an agent to diagnose and propose code changes; it will attempt fixes, run builds, and produce diffs and evidence (logs/screenshots) → review the proposed patch, re-run CI, and approve when the failure is resolved.

3. 3
   Validate an end-to-end web checkout

   Point Antigravity at the web app URL and record the desired checkout path in a short prompt → the agent uses built-in browser automation to click through flows, capture screenshots, and log each step as evidence → export the plan, screenshots, and logs as test artifacts for QA or compliance review.

4. 4
   Run parallel regression tasks nightly

   Schedule multiple agents to run separate regression suites in parallel from the Antigravity UI → each agent persists its artifacts and task-level context for later inspection, including failures and execution trails → review aggregated evidence in the morning to spot flaky UI tests or recurring failures.

5. 5
   Refactor a cross-file API name safely

   Describe the refactor desired (rename function/field across repo) and select a model for the task → Antigravity’s agents plan edits across files, run compile/test steps, and present diffs and execution evidence to confirm behavior → accept or tweak the diff, then merge once tests and traces validate the change.

Pros & Cons

Getting Started

1. 1 Sign in on the Antigravity site and start with the free Individual plan (download the desktop app if preferred).
2. 2 Create a task, select a model (e.g., Gemini 3.1 Pro or Opus 4.6), and point it at your repo or app URL.
3. 3 Run the agent and review the generated plan, code changes, and screenshots within a few minutes.

Pricing