Enable Opencode to authenticate against Antigravity (Google's IDE) via OAuth so you can use Antigravity rate limits and access models like gemini-3-pro and claude-opus-4-5-thinking with your Google credentials.
- Google OAuth sign-in with automatic token refresh via
opencode auth login - Dual Quota System - Access both Antigravity quota (Claude, Gemini 3) and Gemini CLI quota from a single plugin
- Multi-Account Rotation - Add multiple Google accounts; automatically rotates when one is rate-limited
- Real-time SSE streaming including thinking blocks and incremental output
- Model Variants - Configure thinking budget dynamically via OpenCode's variant system
- Extended Thinking - Native support for Claude thinking budgets and Gemini 3 thinking levels
- Auto Recovery - Automatic session recovery from Claude tool_result_missing errors
- Plugin Compatible - Works alongside other OpenCode plugins (opencodesync, etc.)
Option A: Let an LLM do it
Paste this into any LLM agent (Claude Code, OpenCode, Cursor, etc.):
Install the opencode-antigravity-auth plugin and add the Antigravity model definitions to ~/.config/opencode/opencode.json by following: https://raw.githubusercontent.com/NoeFabris/opencode-antigravity-auth/dev/README.md
Option B: Manual setup
-
Add the plugin to your config (
~/.config/opencode/opencode.json):{ "plugin": ["opencode-antigravity-auth@beta"] } -
Authenticate:
opencode auth login
-
Add models (see Available Models for full list):
{ "plugin": ["opencode-antigravity-auth@beta"], "provider": { "google": { "models": { "antigravity-claude-sonnet-4-5-thinking": { "name": "Claude Sonnet 4.5 Thinking", "limit": { "context": 200000, "output": 64000 }, "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }, "variants": { "low": { "thinkingConfig": { "thinkingBudget": 8192 } }, "max": { "thinkingConfig": { "thinkingBudget": 32768 } } } } } } } } -
Use it:
opencode run "Hello" --model=google/antigravity-claude-sonnet-4-5-thinking --variant=max
Installation Guide for LLM Agents
-
Edit the OpenCode configuration file:
- Linux/Mac:
~/.config/opencode/opencode.json - Windows:
%APPDATA%\opencode\opencode.json
- Linux/Mac:
-
Add the plugin to the
pluginsarray -
Set
providerto"google"and choose a model
Create ~/.config/opencode/opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"plugin": ["opencode-antigravity-auth@beta"],
"provider": {
"google": {
"models": {
"antigravity-gemini-3-pro": {
"name": "Gemini 3 Pro (Antigravity)",
"limit": { "context": 1048576, "output": 65535 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingLevel": "low" },
"high": { "thinkingLevel": "high" }
}
},
"antigravity-gemini-3-flash": {
"name": "Gemini 3 Flash (Antigravity)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"minimal": { "thinkingLevel": "minimal" },
"low": { "thinkingLevel": "low" },
"medium": { "thinkingLevel": "medium" },
"high": { "thinkingLevel": "high" }
}
},
"antigravity-claude-sonnet-4-5": {
"name": "Claude Sonnet 4.5 (Antigravity)",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"antigravity-claude-sonnet-4-5-thinking": {
"name": "Claude Sonnet 4.5 Thinking (Antigravity)",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingConfig": { "thinkingBudget": 8192 } },
"max": { "thinkingConfig": { "thinkingBudget": 32768 } }
}
},
"antigravity-claude-opus-4-5-thinking": {
"name": "Claude Opus 4.5 Thinking (Antigravity)",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingConfig": { "thinkingBudget": 8192 } },
"max": { "thinkingConfig": { "thinkingBudget": 32768 } }
}
},
"gemini-2.5-flash": {
"name": "Gemini 2.5 Flash (Gemini CLI)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"gemini-2.5-pro": {
"name": "Gemini 2.5 Pro (Gemini CLI)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"gemini-3-flash-preview": {
"name": "Gemini 3 Flash Preview (Gemini CLI)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"gemini-3-pro-preview": {
"name": "Gemini 3 Pro Preview (Gemini CLI)",
"limit": { "context": 1048576, "output": 65535 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
}
}
}
}
}opencode run "Hello" --model=google/antigravity-claude-sonnet-4-5-thinking --variant=maxModels with antigravity- prefix use Antigravity quota. Thinking models support variants for dynamic thinking budget configuration:
| Model | Variants | Description |
|---|---|---|
google/antigravity-gemini-3-pro |
low, high | Gemini 3 Pro with configurable thinking |
google/antigravity-gemini-3-flash |
minimal, low, medium, high | Gemini 3 Flash with configurable thinking |
google/antigravity-claude-sonnet-4-5 |
- | Claude Sonnet 4.5 (no thinking) |
google/antigravity-claude-sonnet-4-5-thinking |
low, max | Claude Sonnet with configurable thinking |
google/antigravity-claude-opus-4-5-thinking |
low, max | Claude Opus with configurable thinking |
Variant configuration:
- Gemini 3: Uses
thinkingLevelstring ("low","medium","high") - Claude: Uses
thinkingBudgetnumber (8192, 32768 tokens)
Usage:
opencode run "Hello" --model=google/antigravity-claude-sonnet-4-5-thinking --variant=maxModels with -preview suffix use Gemini CLI quota:
| Model | Description |
|---|---|
google/gemini-2.5-flash |
Gemini 2.5 Flash |
google/gemini-2.5-pro |
Gemini 2.5 Pro |
google/gemini-3-flash-preview |
Gemini 3 Flash (preview) |
google/gemini-3-pro-preview |
Gemini 3 Pro (preview) |
Full models configuration (copy-paste ready)
{
"$schema": "https://opencode.ai/config.json",
"plugin": ["opencode-antigravity-auth@beta"],
"provider": {
"google": {
"models": {
"antigravity-gemini-3-pro": {
"name": "Gemini 3 Pro (Antigravity)",
"limit": { "context": 1048576, "output": 65535 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingLevel": "low" },
"high": { "thinkingLevel": "high" }
}
},
"antigravity-gemini-3-flash": {
"name": "Gemini 3 Flash (Antigravity)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"minimal": { "thinkingLevel": "minimal" },
"low": { "thinkingLevel": "low" },
"medium": { "thinkingLevel": "medium" },
"high": { "thinkingLevel": "high" }
}
},
"antigravity-claude-sonnet-4-5": {
"name": "Claude Sonnet 4.5 (no thinking) (Antigravity)",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"antigravity-claude-sonnet-4-5-thinking": {
"name": "Claude Sonnet 4.5 Thinking (Antigravity)",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingConfig": { "thinkingBudget": 8192 } },
"max": { "thinkingConfig": { "thinkingBudget": 32768 } }
}
},
"antigravity-claude-opus-4-5-thinking": {
"name": "Claude Opus 4.5 Thinking (Antigravity)",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingConfig": { "thinkingBudget": 8192 } },
"max": { "thinkingConfig": { "thinkingBudget": 32768 } }
}
},
"gemini-2.5-flash": {
"name": "Gemini 2.5 Flash (Gemini CLI)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"gemini-2.5-pro": {
"name": "Gemini 2.5 Pro (Gemini CLI)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"gemini-3-flash-preview": {
"name": "Gemini 3 Flash Preview (Gemini CLI)",
"limit": { "context": 1048576, "output": 65536 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
},
"gemini-3-pro-preview": {
"name": "Gemini 3 Pro Preview (Gemini CLI)",
"limit": { "context": 1048576, "output": 65535 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
}
}
}
}
}OpenCode's variant system allows you to configure thinking budget dynamically instead of defining separate models for each thinking level.
When you define a model with variants, OpenCode will show variant options in the model picker. Selecting a variant passes the providerOptions to the plugin, which extracts the thinking configuration.
{
"antigravity-claude-sonnet-4-5-thinking": {
"name": "Claude Sonnet 4.5 Thinking",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingConfig": { "thinkingBudget": 8192 } },
"max": { "thinkingConfig": { "thinkingBudget": 32768 } }
}
}
}The plugin accepts different variant formats depending on the model family:
| Model Family | Variant Format |
|---|---|
| Gemini 3 | { "thinkingLevel": "low" | "medium" | "high" } |
| Claude | { "thinkingConfig": { "thinkingBudget": N } } |
| Gemini 2.5 | { "thinkingConfig": { "thinkingBudget": N } } |
Gemini 3 models use string-based thinking levels. Available levels differ by model:
| Level | Flash | Pro | Description |
|---|---|---|---|
minimal |
Yes | No | Minimal thinking, lowest latency |
low |
Yes | Yes | Light thinking |
medium |
Yes | No | Balanced thinking |
high |
Yes | Yes | Maximum thinking (default for both Pro and Flash) |
Note: The API will reject invalid levels (e.g.,
"minimal"on Pro). Configure variants accordingly.
For Gemini 3 models, the old thinkingBudget format is still supported but deprecated:
| Budget Range | Maps to Level |
|---|---|
| ≤ 8192 | low |
| ≤ 16384 | medium |
| > 16384 | high |
Legacy tier-suffixed models still work:
antigravity-claude-sonnet-4-5-thinking-lowantigravity-claude-sonnet-4-5-thinking-mediumantigravity-claude-sonnet-4-5-thinking-highantigravity-gemini-3-pro-lowantigravity-gemini-3-pro-high
However, we recommend using variants for a cleaner model picker and more flexibility.
Add multiple Google accounts for higher combined quotas. The plugin automatically rotates between accounts when one is rate-limited.
opencode auth loginHow multi-account works
- Sticky account selection - Sticks to the same account until rate-limited (preserves Anthropic's prompt cache)
- Per-model-family limits - Rate limits tracked separately for Claude and Gemini models
- Dual quota pools for Gemini - Automatic fallback between Antigravity quota and Gemini CLI quota before switching accounts
- Smart retry threshold - Short rate limits (≤5s) are retried on same account
- Exponential backoff - Increasing delays for consecutive rate limits
For Gemini models, the plugin accesses two independent quota pools per account:
| Quota Pool | When Used |
|---|---|
| Antigravity | Primary (tried first) |
| Gemini CLI | Fallback when Antigravity is rate-limited |
This effectively doubles your Gemini quota per account.
When running opencode auth login with existing accounts:
2 account(s) saved:
1. user1@gmail.com
2. user2@gmail.com
(a)dd new account(s) or (f)resh start? [a/f]:
- Stored in
~/.config/opencode/antigravity-accounts.json - Contains OAuth refresh tokens - treat like a password
- If Google revokes a token (
invalid_grant), that account is automatically removed
Create ~/.config/opencode/antigravity.json (or .opencode/antigravity.json in project root):
| Option | Default | Description |
|---|---|---|
quiet_mode |
false |
Suppress toast notifications (except recovery) |
debug |
false |
Enable debug logging to file |
log_dir |
OS default | Custom directory for debug logs |
auto_update |
true |
Enable automatic plugin updates |
| Option | Default | Description |
|---|---|---|
session_recovery |
true |
Auto-recover from tool_result_missing errors |
auto_resume |
true |
Auto-send resume prompt after recovery |
resume_text |
"continue" |
Text to send when auto-resuming |
| Option | Default | Description |
|---|---|---|
empty_response_max_attempts |
4 |
Retries for empty API responses |
empty_response_retry_delay_ms |
2000 |
Delay between retries |
tool_id_recovery |
true |
Fix mismatched tool IDs from context compaction |
claude_tool_hardening |
true |
Prevent tool parameter hallucination |
| Option | Default | Description |
|---|---|---|
proactive_token_refresh |
true |
Refresh tokens before expiry |
proactive_refresh_buffer_seconds |
1800 |
Refresh 30min before expiry |
max_rate_limit_wait_seconds |
300 |
Max wait time when rate limited (0=unlimited) |
quota_fallback |
false |
Try alternate quota when rate limited |
| Option | Default | Description |
|---|---|---|
account_selection_strategy |
"sticky" |
Strategy for distributing requests across accounts |
pid_offset_enabled |
false |
Use PID-based offset for multi-session distribution |
Available strategies:
| Strategy | Behavior | Best For |
|---|---|---|
sticky |
Same account until rate-limited | Prompt cache preservation |
round-robin |
Rotate to next account on every request | Maximum throughput |
hybrid |
Touch all fresh accounts first, then sticky | Sync reset timers + cache |
Error handling:
| Error Type | Behavior |
|---|---|
MODEL_CAPACITY_EXHAUSTED |
Wait (escalating 5s→60s) and retry same account |
QUOTA_EXCEEDED |
Switch to next available account immediately |
This prevents unnecessary account switching when server-side capacity issues affect all accounts equally.
OPENCODE_ANTIGRAVITY_QUIET=1 # quiet_mode
OPENCODE_ANTIGRAVITY_DEBUG=1 # debug
OPENCODE_ANTIGRAVITY_LOG_DIR=/path # log_dir
OPENCODE_ANTIGRAVITY_KEEP_THINKING=1 # keep_thinking
OPENCODE_ANTIGRAVITY_ACCOUNT_SELECTION_STRATEGY=round-robin # account_selection_strategy
OPENCODE_ANTIGRAVITY_PID_OFFSET_ENABLED=1 # pid_offset_enabledFull configuration example
{
"$schema": "https://raw.githubusercontent.com/NoeFabris/opencode-antigravity-auth/main/assets/antigravity.schema.json",
"quiet_mode": false,
"debug": false,
"log_dir": "/custom/log/path",
"auto_update": true,
"keep_thinking": false,
"session_recovery": true,
"auto_resume": true,
"resume_text": "continue",
"empty_response_max_attempts": 4,
"empty_response_retry_delay_ms": 2000,
"tool_id_recovery": true,
"claude_tool_hardening": true,
"proactive_token_refresh": true,
"proactive_refresh_buffer_seconds": 1800,
"proactive_refresh_check_interval_seconds": 300,
"max_rate_limit_wait_seconds": 300,
"quota_fallback": false,
"account_selection_strategy": "sticky",
"pid_offset_enabled": false,
"signature_cache": {
"enabled": true,
"memory_ttl_seconds": 3600,
"disk_ttl_seconds": 172800,
"write_interval_seconds": 60
}
}If you encounter auth issue please try remove antigravity-account.json and auth again
Try add this line to in google field under provider
"npm": "@ai-sdk/google"
If you encounter error during the session, try chat continue the recover session mechanism should be trigger and you can continue the session, if the error blocked the session please workaround by use command /undo to revert to the state before the error and try again it should work
DCP creates synthetic assistant messages that lack thinking blocks. Our plugin must be listed BEFORE DCP:
{
"plugin": [
"opencode-antigravity-auth@beta",
"@tarquinen/opencode-dcp@latest",
]
}When using opencode-antigravity-auth, disable the built-in auth and override agent models in oh-my-opencode.json:
{
"google_auth": false,
"agents": {
"frontend-ui-ux-engineer": { "model": "google/gemini-3-pro-high" },
"document-writer": { "model": "google/gemini-3-flash" },
"multimodal-looker": { "model": "google/gemini-3-flash" }
}
}When spawning parallel subagents, multiple processes may select the same account causing rate limit errors. Workaround: Enable pid_offset_enabled: true to distribute sessions across accounts, or add more accounts via opencode auth login.
- gemini-auth plugins - Not needed. This plugin handles all Google OAuth authentication.
Migration Guide (v1.2.8+ - Variants)
v1.2.8+ introduces model variants for dynamic thinking configuration. Instead of separate models for each thinking level, you now define one model with variants.
{
"antigravity-claude-sonnet-4-5-thinking-low": { ... },
"antigravity-claude-sonnet-4-5-thinking-medium": { ... },
"antigravity-claude-sonnet-4-5-thinking-high": { ... }
}{
"antigravity-claude-sonnet-4-5-thinking": {
"name": "Claude Sonnet 4.5 Thinking",
"limit": { "context": 200000, "output": 64000 },
"modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
"variants": {
"low": { "thinkingConfig": { "thinkingBudget": 8192 } },
"max": { "thinkingConfig": { "thinkingBudget": 32768 } }
}
}
}- Cleaner model picker - 4 models instead of 12+
- Flexible budgets - Define any budget, not just low/max
- Future-proof - Works with OpenCode's native variant system
Old tier-suffixed models (antigravity-claude-sonnet-4-5-thinking-low, etc.) still work. No action required if you prefer the old style.
Migration Guide (v1.2.7 - Prefix)
If upgrading from v1.2.6 or earlier:
v1.2.7+ uses explicit prefixes to distinguish quota sources:
| Model Type | New Name (Recommended) | Old Name (Still Works) |
|---|---|---|
| Gemini 3 (Antigravity) | antigravity-gemini-3-pro |
gemini-3-pro-low |
| Gemini 3 (CLI) | gemini-3-pro-preview |
N/A |
| Claude | antigravity-claude-sonnet-4-5 |
claude-sonnet-4-5 |
Update your config to use antigravity- prefix:
- "gemini-3-pro-low": { ... }
+ "antigravity-gemini-3-pro": { ... }Why update? Old names work now as a fallback, but this depends on Gemini CLI using
-previewsuffix. If Google removes-previewin the future, old names may route to the wrong quota. Theantigravity-prefix is explicit and stable.
rm -rf ~/.config/opencode/antigravity-account.json
opencode auth loginModels now use antigravity- prefix for Antigravity quota. See Available Models.
{
"$schema": "https://raw.githubusercontent.com/NoeFabris/opencode-antigravity-auth/main/assets/antigravity.schema.json",
"quiet_mode": false,
"debug": false
}E2E Testing
The plugin includes regression tests for stability verification. Tests consume API quota.
# Sanity tests (7 tests, ~5 min)
npx tsx script/test-regression.ts --sanity
# Heavy tests (4 tests, ~30 min)
npx tsx script/test-regression.ts --heavy
# Concurrent tests (3 tests)
npx tsx script/test-regression.ts --category concurrency
# Run specific test
npx tsx script/test-regression.ts --test thinking-tool-use
# List tests without running
npx tsx script/test-regression.ts --dry-runOPENCODE_ANTIGRAVITY_DEBUG=1 opencode # Basic logging
OPENCODE_ANTIGRAVITY_DEBUG=2 opencode # Verbose (full request/response bodies)Logs are written to ~/.config/opencode/antigravity-logs/.
- Architecture - Plugin internals and request flow
- API Spec - Antigravity API reference
Safety, Usage & Legal
- Personal / internal development only
- Respect internal quotas and data handling policies
- Not for production services or bypassing intended limits
By using this plugin, you acknowledge:
- Terms of Service risk - This approach may violate ToS of AI model providers
- Account risk - Providers may suspend or ban accounts
- No guarantees - APIs may change without notice
- Assumption of risk - You assume all legal, financial, and technical risks
- Not affiliated with Google. This is an independent open-source project.
- "Antigravity", "Gemini", "Google Cloud", and "Google" are trademarks of Google LLC.
- Software is provided "as is", without warranty.
Built with help from:
- opencode-gemini-auth - Gemini OAuth groundwork by @jenslys
- CLIProxyAPI - Antigravity API reference
If this plugin helps you, consider supporting its continued maintenance:
MIT
![@github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=64&v=4){kind=small}
![@coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=64&v=4){kind=small}
