pi-antigravity
September 4, 2026 · View on GitHub
pi-antigravity is a Pi Coding Agent provider that lets Pi talk directly to Google Antigravity / Cloud Code Assist models — Gemini, plus the Claude and GPT-OSS models Antigravity also advertises. Sign in with Google, pick a model, and go. Under the hood it handles OAuth login, native streaming, model routing, and quota diagnostics itself, so it never shells out to an external Antigravity CLI.
Using OpenCode instead of Pi? Install the companion plugin @rahularya01/opencode-antigravity.
Unofficial integration. This project is not affiliated with or endorsed by Google. Use it only with an account and services you are authorized to access, and review its source before granting OAuth permissions.
Contents
- Requirements
- Install
- Quick start
- Authentication and credential safety
- Commands
- Models and routing
- Configuration
- Troubleshooting
- Development
Requirements
- Pi Coding Agent and Pi AI version 0.80.0 or later
- A Google account that can use the relevant Cloud Code Assist / Antigravity services
- A browser to complete the Google sign-in. Same-machine is best (the browser hits the local callback automatically); on a remote/headless machine, complete sign-in anywhere and paste the resulting callback URL back into Pi (see Troubleshooting).
Install
Install from npm:
pi install npm:pi-antigravity
Or install the latest repository version:
pi install git:github.com/Rahularya01/pi-antigravity
Restart Pi (or run /reload) after installation. To update the npm package later, use pi update npm:pi-antigravity.
Quick start
-
Start Pi and run
/login antigravity. -
Complete Google sign-in in your browser.
-
Select a model, for example:
/model antigravity/gemini-3.8-flash -
Start working. If a request fails, run
/antigravity.doctorfor sanitized diagnostics.
Authentication and credential safety
The provider uses the OAuth 2.0 Authorization Code flow with PKCE, so credentials are only ever exchanged with Google — never typed into Pi.
/login antigravityopens Google sign-in and starts a temporary callback listener athttp://localhost:51121/oauth-callback.- After you approve access, Pi exchanges the callback code for tokens and stores the provider credentials in Pi's auth store (normally
~/.pi/agent/auth.json). - Pi refreshes access tokens automatically when they expire — you shouldn't need to sign in again unless a token is revoked.
The callback listener binds only to a loopback host, so it isn't reachable from outside your machine. The auth file it writes to contains sensitive access and refresh tokens: do not commit it, paste it into issues, or share its contents.
Signing in requests these Google OAuth scopes:
| Scope | Why it's needed |
|---|---|
aicode | Access to the Cloud Code Assist / Antigravity model catalog and endpoints |
cloud-platform | General Cloud Code Assist API access |
userinfo.email, userinfo.profile | Identify the signed-in Google account |
cclog | Cloud Code Assist logging/telemetry endpoints used by the API |
experimentsandconfigs | Server-side experiment and config flags for the API |
Review these permissions before approving access. If your credentials expire or are revoked, just re-run /login antigravity to sign in again.
Commands
| Command | Description |
|---|---|
/login antigravity | Sign in to Google and configure the provider. |
/model antigravity/<model-id> | Choose a registered Antigravity model. |
/antigravity.usage | Show the server-reported shared quota groups and reset times. |
/antigravity.models | List available runtime models, remaining shared-pool quota, and capabilities. |
/antigravity.models all | Include tab/chat models normally hidden from the model list. |
/antigravity.refresh | Force refresh the dynamic model catalog from Antigravity. |
/antigravity.doctor | Show sanitized provider diagnostics, including the endpoint, status, and resolved runtime model. |
/antigravity.image <prompt> | Generate an image via Antigravity and save it under .pi/generated-images/. Optional --ratio 16:9, --model, --path. |
Model availability, entitlement, quota groups, and resets are returned by the service and can differ by account. The quota percentage shown for a model can represent a shared pool, not a private per-model allowance.
The extension also registers a generate_image tool the model can call. Images are written inside the project directory (default .pi/generated-images/). Image models such as gemini-3-pro-image are account-dependent; /antigravity.image falls back to other advertised Gemini image IDs on 404.
Models and routing
After you sign in, the provider refreshes its catalog from Antigravity (fetchAvailableModels) and groups runtime thinking variants into public Pi model IDs. Newly enabled models — for example a new Gemini Flash generation — become selectable after that refresh without waiting for an extension release. A last-known-good cache is kept for offline/cold start; the static table below is only the conservative fallback and a routing reference.
Use /antigravity.models to see live availability and quota for your account. Runtime names such as gemini-3.8-flash-low / -medium / -high collapse to gemini-3.8-flash with those thinking levels. The conservative static entries remain selectable when an account's authenticated catalog omits them.
Why Claude and GPT-OSS appear
Antigravity / Cloud Code Assist exposes a multi-provider catalog. Depending on your account, its Google-authenticated API can advertise Google Gemini models alongside Claude models served through Anthropic Vertex and GPT-OSS served through OpenAI Vertex. This extension intentionally exposes those advertised Claude and GPT-OSS models through the single antigravity provider; they are not separate Pi providers and do not use a separate Anthropic or OpenAI login.
The backend's display labels do not always match its runtime IDs. For example, gemini-3.5-flash-extra-low, gemini-3.5-flash-low, and gemini-3-flash-agent can be displayed as Gemini 3.5 Flash Low, Medium, and High. Gemini 3.8, 3.7, and 3.6 Flash use per-effort runtime IDs and send thinkingLevel; Gemini 3.5 Flash and 3.1 Pro send thinkingBudget.
| Public model ID | Input | Thinking levels shown | Max output tokens | Request routing |
|---|---|---|---|---|
gemini-3.8-flash | Text, image | Low, Medium, High | 65,536 | low → gemini-3.8-flash-low; medium → gemini-3.8-flash-medium; high → gemini-3.8-flash-high |
gemini-3.7-flash | Text, image | Low, Medium, High | 65,536 | low → gemini-3.7-flash-low; medium → gemini-3.7-flash-medium; high → gemini-3.7-flash-high |
gemini-3.6-flash | Text, image | Low, Medium, High | 65,536 | low → gemini-3.6-flash-low; medium → gemini-3.6-flash-medium; high → gemini-3.6-flash-high |
gemini-3.5-flash | Text, image | Low, Medium, High | 65,536 | low → gemini-3.5-flash-extra-low; medium → gemini-3.5-flash-low; high → gemini-3-flash-agent |
gemini-3.1-pro | Text, image | Low, High | 65,535 | low → gemini-3.1-pro-low; high → gemini-pro-agent |
claude-sonnet-4-6 | Text, image | High | 64,000 | high → claude-sonnet-4-6 |
claude-opus-4-6 | Text, image | High | 64,000 | high → claude-opus-4-6-thinking |
gpt-oss-120b | Text | Medium | 32,768 | medium → gpt-oss-120b-medium |
To limit which models Pi cycles through, enable specific entries in ~/.pi/agent/settings.json:
{
"models": {
"antigravity/gemini-3.8-flash": { "enabled": true },
"antigravity/gemini-3.7-flash": { "enabled": true },
"antigravity/gemini-3.6-flash": { "enabled": true },
"antigravity/gemini-3.5-flash": { "enabled": true },
"antigravity/gemini-3.1-pro": { "enabled": true },
"antigravity/claude-sonnet-4-6": { "enabled": true }
}
}
Configuration
All primary environment variables start with ANTIGRAVITY_. The legacy NOAGY_ prefix is also accepted for compatibility.
| Variable | Purpose |
|---|---|
ANTIGRAVITY_BASE_URL | Override the API base URL. It must be HTTPS, contain no URL credentials, and target an allowed Google APIs host. |
ANTIGRAVITY_PROJECT_ID | Use a specific Cloud Code Assist project ID instead of discovery or the stable account fallback. |
ANTIGRAVITY_CALLBACK_HOST | Bind OAuth callback to 127.0.0.1, ::1, or localhost only. Defaults to 127.0.0.1. |
ANTIGRAVITY_USER_AGENT | Override the request user-agent. |
ANTIGRAVITY_RUNTIME_MODEL | Pin requests to a runtime model ID, bypassing discovered/fallback routing. |
ANTIGRAVITY_CATALOG_REFRESH_INTERVAL_MS | Override catalog refresh TTL in ms. Defaults to 4 hours (14400000). Set to 0 to always refresh. |
ANTIGRAVITY_CLIENT_ID | Use a custom Google OAuth client ID. |
ANTIGRAVITY_CLIENT_SECRET | Use a custom Google OAuth client secret. Keep it out of source control and shell history. |
ANTIGRAVITY_NO_KEEPALIVE | Set to 1 to skip the keep-alive connection pool. |
ANTIGRAVITY_NO_PREWARM | Set to 1 to skip the TLS pre-warm request made when the extension loads. |
By default, the provider tries https://daily-cloudcode-pa.googleapis.com, then the sandbox host, then https://cloudcode-pa.googleapis.com. Prefer the built-in OAuth client unless you have a reason to use your own credentials.
Latency
Provider requests reuse a keep-alive connection pool when the runtime supports it, so consecutive turns do not repeat the DNS, TCP, and TLS handshake. When HTTP_PROXY, HTTPS_PROXY, or ALL_PROXY is set, that pool is skipped so Pi's proxy-aware dispatcher is used instead. The connection is also opened when the extension loads so the first message of a session skips the handshake too. For the lowest time-to-first-token, pick a fast runtime: gemini-3.8-flash with reasoning off routes to gemini-3.8-flash-low at thinking level LOW. Setting ANTIGRAVITY_PROJECT_ID also removes the project-discovery round-trip when credentials do not already carry a project ID.
Troubleshooting
- No credentials / 401 / 403: Run
/login antigravityagain, then check/antigravity.doctor. - Remote/headless machine — browser can't reach
localhost:51121: The callback binds to loopback only, so a browser on another machine can't hit it. You have two options:- Paste (no extra setup): Run
/login antigravity, open the shown URL and complete Google sign-in in any browser. When it redirects tohttp://localhost:51121/oauth-callback?…and fails to load, copy that full URL from the address bar and paste it into the prompt Pi shows. The code is single-use and expires quickly, so paste promptly. - SSH tunnel (reusable): From the machine with the browser, run
ssh -N -L 51121:127.0.0.1:51121 <user>@<server>and keep it open, then run/login antigravityon the server. The redirect tolocalhost:51121tunnels through to the local callback automatically.
- Paste (no extra setup): Run
- OAuth callback will not start: Ensure port
51121is free andANTIGRAVITY_CALLBACK_HOSTis a permitted loopback address. - Model is unavailable: Run
/antigravity.models; availability is account- and service-dependent. - Claude/GPT tool-call schema error: Upgrade to the latest package release. The provider adapts Pi's JSON Schema tool definitions for the Cloud Code Assist custom-tool bridge.
- Quota or rate limit: Run
/antigravity.usage. A429response usually indicates quota or rate limiting; changing models may still draw from the same shared pool. - Need a safe diagnostic:
/antigravity.doctorredacts recognized secrets from its error output. Still review output before sharing it publicly.
Development
This repo uses Bun for install, scripts, and CI. The published extension itself runs on Node (Pi's CLI).
bun install
bun run check
The package declares its Pi extension in package.json under pi.extensions. See the Pi package documentation for package installation, manifest, and gallery conventions.
Support the project
If pi-antigravity is useful to you, consider sponsoring the project on GitHub.