> ## Documentation Index
> Fetch the complete documentation index at: https://docs.replicas.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Gemini

> Run Google's Antigravity CLI with your Google account in Replicas workspaces.

Replicas runs Google's Antigravity CLI (`agy`) in the workspace over the Agent Client Protocol.

## Setup

Gemini signs in with a Google account that has an Antigravity or Google AI subscription.

Open **Gemini** in [Organization → Coding Agents](https://app.replicas.dev/dashboard/agents) or [Personal → Coding Agents](https://app.replicas.dev/dashboard/account/agents) and click **Sign in with Google**. After you sign in, the callback page shows a code; paste it back into Replicas.

Or connect from the [Replicas CLI](https://docs.replicas.dev/features/cli.md):

```bash theme={null}
replicas login
replicas gemini-auth
```

The credential is personal by default. Run `replicas gemini-auth --org` as an organization admin to share it with your organization. See [Credentials](https://docs.replicas.dev/admin/credentials.md) for scope and precedence.

## Models

| Model ID | Label | Notes |
| - | - | - |
| `gemini-3.8-flash` | Gemini 3.8 Flash | Default, every plan |
| `gemini-4-neon` | Gemini 4 Neon | Google AI Pro or Ultra |
| `gemini-4-carbon` | Gemini 4 Carbon | Google AI Ultra |
| `gemini-4-argon` | Gemini 4 Argon | Google AI Pro or Ultra |
| `gemini-4-hydrogen` | Gemini 4 Hydrogen | Google AI Pro or Ultra |
| `gemini-pro-orchestrator-neon` | Orchestrator (Neon) | Plans with Gemini 4 Neon, runs with Gemini 4 Hydrogen. Google AI Pro or Ultra |
| `gemini-pro-orchestrator-g1-pro` | Orchestrator (Argon) | Plans with Gemini 4 Argon, runs with Gemini 4 Hydrogen. Google AI Pro or Ultra |
| `claude-opus-5-5` | Claude Opus 5.5 (Thinking) | Google AI Pro or Ultra |
| `claude-sonnet-5-5` | Claude Sonnet 5.5 (Thinking) | Google AI Pro or Ultra |
| `claude-opus-4-6-thinking` | Claude Opus 4.6 (Thinking) | Free and Pro plans |
| `claude-sonnet-4-6` | Claude Sonnet 4.6 | Free and Pro plans |
| `gpt-oss-120b-medium` | GPT-OSS 120B | Every plan |

Which models work depends on your [Antigravity plan](https://antigravity.google/pricing); picking one your plan lacks lists the models you have. Effort levels `low` through `max` set Antigravity's thinking level.

## Using Gemini

Choose **Gemini** from the workspace agent picker. Gemini supports your Antigravity slash commands and [skills](https://docs.replicas.dev/features/environments.md), questions you answer in chat, image input, context usage, [MCP servers](https://docs.replicas.dev/features/mcp.md), and Replicas command protection. Fast mode and goals are not supported.

If it reaches a usage limit, Replicas can fork the chat to another harness. See [Fallback chains](https://docs.replicas.dev/admin/credentials.md#fallback-chains).

## Programmatic use

Use `gemini` anywhere Replicas accepts a coding-agent provider:

```bash theme={null}
replicas create fix-tests -e backend -a gemini -m "Fix the failing tests"
replicas send REPLICA_ID -a gemini -m "Run the tests again"
```

The [API](https://docs.replicas.dev/features/api.md), [MCP](https://docs.replicas.dev/features/mcp.md), [Automations](https://docs.replicas.dev/features/automations.md), and [Defaults](https://docs.replicas.dev/admin/defaults.md) surfaces also accept `gemini`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.