Gemini OAuth Plugin for Opencode

[!WARNING] Google has stated that using Gemini CLI OAuth with third-party software is a policy-violating use case and may trigger abuse detection or account restrictions. It is unclear how aggressively this is enforced for projects like this one, but you should assume there is real risk and use this plugin at your own discretion. If you want the lowest-risk option, use Opencode with your own Gemini API key instead.

See: https://github.com/google-gemini/gemini-cli/discussions/22970

Authenticate the Opencode CLI with your Google account. This plugin enables you to use your existing Gemini plan and quotas (including the free tier) directly within Opencode.

Prerequisites

• Opencode CLI installed.
• A Google account with access to Gemini.

Installation

Add the plugin to your Opencode configuration file (~/.config/opencode/opencode.json or similar):

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-gemini-auth@latest"]
}

[!IMPORTANT] Explicitly configure a Google Cloud projectId if you're using an organization-backed Gemini Code Assist subscription (Standard/Enterprise) or a company, school, or Google Workspace account. Most individual Google accounts should not need this. Google AI Plus is not a Gemini Code Assist subscription tier. You can still set projectId to force a specific project.

Usage

1. Login: Run the authentication command in your terminal:

   opencode auth login

2. Select Provider: Choose Google from the list.

3. Authenticate: Select OAuth with Google (Gemini CLI).

   • A browser window will open for you to approve the access.
   • The plugin spins up a temporary local server to capture the callback.
   • If the local server fails (e.g., port in use or headless environment), you can manually paste the callback URL or just the authorization code.

Once authenticated, Opencode will use your Google account for Gemini requests.

To check your current Gemini Code Assist quota buckets at any time, run:

/gquota

Configuration

Google Cloud Project

By default, the plugin attempts to provision or find a suitable Google Cloud project. To force a specific project, set the projectId in your configuration or via environment variables:

File: ~/.config/opencode/opencode.json

{
  "provider": {
    "google": {
      "options": {
        "projectId": "your-specific-project-id"
      }
    }
  }
}

You can also set OPENCODE_GEMINI_PROJECT_ID, GOOGLE_CLOUD_PROJECT, or GOOGLE_CLOUD_PROJECT_ID to supply the project ID via environment variables.

Proxy

If your network requires an HTTP proxy for Google API calls, set OPENCODE_GEMINI_AUTH_PROXY before starting Opencode:

OPENCODE_GEMINI_AUTH_PROXY=http://127.0.0.1:8080 opencode

This is passed to Bun's fetch proxy option and applies to OAuth, token refresh, project/quota lookup, and Gemini request forwarding.

Model list

If you want to remove unusable models from the picker, use OpenCode's provider.google.whitelist or provider.google.blacklist settings.

• whitelist: only show the listed model IDs.
• blacklist: hide specific model IDs from the default list.
• models: define or override model metadata/options, but does not remove the default models by itself.

Use the exact model IDs reported by opencode models google when building these lists.

Example: keep only a small Gemini model list visible.

{
  "provider": {
    "google": {
      "whitelist": [
        "gemini-2.5-flash",
        "gemini-2.5-pro",
        "gemini-3-flash-preview",
        "gemini-3-pro-preview"
      ]
    }
  }
}

Example: hide a few unwanted defaults while keeping the rest.

{
  "provider": {
    "google": {
      "blacklist": [
        "gemini-2.0-flash-exp",
        "gemini-1.5-pro"
      ]
    }
  }
}

Below are example model entries you can add under provider.google.models in your Opencode config. Each model can include an options.thinkingConfig block to enable "thinking" features.

{
  "provider": {
    "google": {
      "models": {
        "gemini-2.5-flash": {
          "options": {
            "thinkingConfig": {
              "thinkingBudget": 8192,
              "includeThoughts": true
            }
          }
        },
        "gemini-2.5-pro": {
          "options": {
            "thinkingConfig": {
              "thinkingBudget": 8192,
              "includeThoughts": true
            }
          }
        },
        "gemini-3-flash-preview": {
          "options": {
            "thinkingConfig": {
              "thinkingLevel": "high",
              "includeThoughts": true
            }
          }
        },
        "gemini-3-pro-preview": {
          "options": {
            "thinkingConfig": {
              "thinkingLevel": "high",
              "includeThoughts": true
            }
          }
        }
      }
    }
  }
}

Note: Available model names and previews may change—check Google's documentation or the Gemini product page for the current model identifiers.

Thinking Models

The plugin supports configuring Gemini "thinking" features per-model via thinkingConfig. The available fields depend on the model family:

• For Gemini 3 models: use thinkingLevel with values "low" or "high".
• For Gemini 2.5 models: use thinkingBudget (token count).
• includeThoughts (boolean) controls whether the model emits internal thoughts.

A combined example showing both model types:

{
  "provider": {
    "google": {
      "models": {
        "gemini-3-pro-preview": {
          "options": {
            "thinkingConfig": {
              "thinkingLevel": "high",
              "includeThoughts": true
            }
          }
        },
        "gemini-2.5-flash": {
          "options": {
            "thinkingConfig": {
              "thinkingBudget": 8192,
              "includeThoughts": true
            }
          }
        }
      }
    }
  }
}

If you don't set a thinkingConfig for a model, the plugin will use default behavior for that model.

The plugin also accepts request payloads that put thinkingConfig at the root and normalizes them into generationConfig.thinkingConfig before forwarding to Gemini Code Assist.

Troubleshooting

Manual Google Cloud Setup

If automatic provisioning fails, you may need to set up the project manually:

1. Go to the Google Cloud Console.
2. Create or select a project.
3. Enable the Gemini for Google Cloud API (cloudaicompanion.googleapis.com).
4. Configure the projectId in your Opencode config as shown above.

Quotas, Plans, and 429 Errors

Common causes of 429 RESOURCE_EXHAUSTED or QUOTA_EXHAUSTED:

• No project ID configured: the plugin uses a managed free-tier project, which has lower quotas.
• Model-specific limits: quotas are tracked per model (e.g., gemini-3-pro-preview vs gemini-3-flash-preview).
• Large prompts: OAuth/Code Assist does not support cached content, so long system prompts and history can burn quota quickly.
• Parallel sessions: multiple Opencode windows can drain the same bucket.

Notes:

• Gemini CLI auto-fallbacks: the official CLI may fall back to Flash when Pro quotas are exhausted, so it can appear to “work” even if the Pro bucket is depleted.
• Org-backed Code Assist subscriptions require a project: if you're using Gemini Code Assist Standard or Enterprise, set provider.google.options.projectId (or OPENCODE_GEMINI_PROJECT_ID) and re-authenticate.

Debugging

To view detailed logs of Gemini requests and responses, set the OPENCODE_GEMINI_DEBUG environment variable:

OPENCODE_GEMINI_DEBUG=1 opencode

This will generate gemini-debug-<timestamp>.log files in your working directory containing sanitized request/response details.

Parity Notes

This plugin mirrors the official Gemini CLI OAuth flow and Code Assist endpoints. In particular, project onboarding and quota retry handling follow the same behavior patterns as the Gemini CLI.

References

• Gemini CLI repository: https://github.com/google-gemini/gemini-cli
• Gemini CLI quota documentation: https://developers.google.com/gemini-code-assist/resources/quotas

Local upstream mirror (optional)

For local parity checks, you can keep a separate clone of the official gemini-cli in this repo at .local/gemini-cli.

This mirror is intentionally untracked, so contributors must set it up once on their machine:

git clone https://github.com/google-gemini/gemini-cli.git .local/gemini-cli

After setup, pull upstream updates with:

bun run update:gemini-cli

Updating

Opencode does not automatically update plugins. To update to the latest version, you must clear the cached plugin:

# Clear the specific plugin cache
rm -rf ~/.cache/opencode/node_modules/opencode-gemini-auth

# Run Opencode to trigger a fresh install
opencode

Development

To develop on this plugin locally:

1. Clone:

   git clone https://github.com/jenslys/opencode-gemini-auth.git
   cd opencode-gemini-auth
   bun install

2. Link: Update your Opencode config to point to your local directory using a file:// URL:

   {
     "plugin": ["file:///absolute/path/to/opencode-gemini-auth"]
   }

License

MIT