Skip to main content

Google Antigravity Integration

Technical guide to governing autonomous and multi-agent environments in Google Antigravity with RAPID OS v3.

1. What Google Antigravity is and how it fits with RAPID OS​

Google Antigravity is an AI-assisted development environment designed for advanced pair programming and the execution of specialized subagents with access to shell tools, file editing, and MCP protocols.

Fit with RAPID OS: Antigravity is a highly capable harness that can read, write, run commands, and delegate to concurrent subagents. RAPID OS provides the deterministic governance structure: immutable contracts, risk gates, and cryptographic verification of the evidence generated by the agents.

2. Configuration in RAPID OS​

RAPID OS includes the built-in antigravity profile:

rapid harness show antigravity

Declared capabilities:

  • context.consume: supported
  • repository.read: supported
  • workspace.current: supported
  • subagents.delegate: available as a canonical governance capability.

3. Configuration in Antigravity​

  1. Configure your workspace and the required MCP servers in Antigravity.
  2. Use project rules to instruct the main agent to always check the state of the RAPID OS contracts before declaring a task complete.
  3. Allow the agent to run tests in controlled terminals.

Main workflow: shortcuts​

The recommended way to work with Antigravity is with the /rapid-<mode> shortcuts. The 9 shortcuts, the loop they follow, human approval, and common errors are in the shortcuts guide.

Install​

rapid init --shortcuts                                # when initializing the project
rapid shortcuts install --harness antigravity         # in an already initialized project

It generates .agents/skills/rapid-*/SKILL.md (the 9 shortcuts; Antigravity skills documentation). It is the same folder Codex uses: if you select both, it is written only once. After updating Rapid OS, run rapid shortcuts install again.

First step: /rapid-init (onboarding)​

The entry point of a project is /rapid-init: it drives the onboarding (rapid setup next), asks you for the stack of each component, the architecture, the services, and the documents, and reviews the drafts with you before closing. After that you use /rapid-feature and the other modes. See First project.

To have /rapid-init and the other shortcuts in all your projects without installing them in each one, install them in your user folder (~/.gemini/config/skills/); the Rapid OS installer asks you about it when it finishes:

rapid shortcuts install --global --harness antigravity

The shortcuts and hooks run rapid from the agent's shell: if it is not on the PATH, rapid init and rapid doctor warn with RAPID1414 (a warning, not an error). See the shortcuts guide.

Invoke​

They are invoked with / in the chat:

/rapid-feature añadir exportación CSV al informe mensual

/rapid-approve does not approve: it shows the pending gates and the evidence, and gives you the command for your terminal. If Antigravity delegates to subagents, the shortcut remains the same: the rapid next → rapid check → rapid finish loop and human approval in your terminal.

Advanced mode (v3 commands)​

The v3 commands (rapid spec, rapid run, rapid evidence, rapid eval…) remain available for CI, debugging, and auditing, and you do not need to run them by hand in day-to-day work either. Section 4 keeps the complete manual workflow.

4. Governing Complex Tasks with Subagents​

In large-scale tasks (such as refactors or hardening), Antigravity can invoke subagents:

  1. Create the Run:
rapid run create --spec spec-complex-refactor --harness antigravity
  1. Verify the Baseline: Run the initial tests, saving the authentic log as an artifact (base.log), record the baseline evidence (as in evidence_baseline.json), and acknowledge gate.baseline before activating the run:
python -m unittest discover -s tests -t . > base.log 2>&1
rapid evidence add --run <run-id> --input evidence_baseline.json
rapid run gate <run-id> gate.baseline acknowledged
rapid run status <run-id> active
rapid run task <run-id> T001 in_progress
  1. Delegation Evidence: If Antigravity delegates a subtask to a specialized subagent, a canonical evidence of type delegation can be recorded:
{
  "kind": "delegation",
  "producer": "antigravity-orchestrator",
  "summary": "Delegación de análisis de dependencias al subagente de arquitectura",
  "task_ids": ["T001"],
  "gate_ids": [],
  "capability_ids": ["subagents.delegate"],
  "payload": {
    "target": "subagent-architect",
    "outcome": "success"
  }
}

  1. Unit Test Evidence: The agent runs the tests in a temporary environment and captures the output with test_result.

  2. Compliance Evaluation:

rapid eval run --run <run-id> --write --require-pass

The evaluator will deterministically verify that all required gates, tasks, and capabilities have been satisfied.