Antigravity Manager - Setup Guide¶
Source: reason-machines/trending-skills (1,396+ installs) Category: Agent Infrastructure / LLM Operations Quality Tier: 🔵 Community
Antigravity Manager is a professional AI account manager and proxy gateway (Tauri v2 + Rust desktop app, plus a headless Docker/server mode). It takes Google (Gemini) and Anthropic (Claude) web session accounts and exposes them as standard OpenAI-compatible, Anthropic-native, and Gemini-native API endpoints with intelligent multi-account rotation, quota tracking, and automatic failover. This skill (from the ara.so Daily 2026 Skills collection) walks an agent through installing, configuring, and operating the gateway.
Installation¶
Step 1 - Install the skill¶
npx skills add reason-machines/trending-skills --skill antigravity-manager
The publisher org was renamed aradotso → reason-machines on Sep 12, 2026. The legacy aradotso/trending-skills install path still resolves (GitHub 301 redirect), but reason-machines is canonical.
Step 2 - Install Antigravity Manager itself¶
| Method | Command |
|---|---|
| One-line installer (Linux/macOS) | curl -fsSL https://raw.githubusercontent.com/lbjlaq/Antigravity-Manager/v4.1.30/install.sh | bash |
| Homebrew | brew tap lbjlaq/antigravity-manager https://github.com/lbjlaq/Antigravity-Manager && brew install --cask antigravity-tools |
| Docker (headless server) | docker run -d --name antigravity-manager -p 8045:8045 -e API_KEY=$ANTIGRAVITY_API_KEY -e WEB_PASSWORD=$ANTIGRAVITY_WEB_PASSWORD -v ~/.antigravity_tools:/root/.antigravity_tools lbjlaq/antigravity-manager:latest |
| Manual | Download .dmg / .msi / .deb / .rpm / .AppImage from GitHub Releases |
Prerequisites¶
| Requirement | Notes |
|---|---|
| Google account(s) with Gemini access | Authorized via the OAuth 2.0 URL generated in the app |
| Anthropic account(s) with Claude access | Same OAuth flow |
| Docker (server mode) or desktop OS | Headless server mode is built from source with --features headless |
API_KEY |
Authenticates proxy API calls |
WEB_PASSWORD (recommended) |
Separate web UI admin password; falls back to API_KEY |
Capabilities¶
| Capability | Trigger / Detail |
|---|---|
| Multi-account rotation | Auto-rotates accounts on 429/401 errors with millisecond failover |
| Protocol conversion | Web sessions → OpenAI (/v1/chat/completions), Anthropic (/v1/messages), Gemini (generateContent) |
| Model routing | Regex remaps in gui_config.json, e.g. gpt-4.* → gemini-2.5-pro |
| Tiered routing | Prioritizes accounts Ultra → Pro → Free; fast-reset accounts consumed first |
| Background-task demotion | Background requests auto-redirected to Flash-tier models to preserve premium quota |
| Quota dashboard | Live Gemini Pro/Flash, Claude, and Imagen 3 quotas with smart recommendation |
| 403 detection | Banned accounts flagged and skipped during rotation |
| Batch account import | JSON import of {token, type, label} records |
API Endpoints¶
The proxy listens on port 8045 by default.
| Protocol | Endpoint | Compatible with |
|---|---|---|
| OpenAI-compatible | POST http://localhost:8045/v1/chat/completions |
Any OpenAI SDK client |
| Anthropic-native | POST http://localhost:8045/v1/messages |
Claude Code |
| Gemini-native | POST http://localhost:8045/v1/models/{model}:generateContent |
Gemini clients |
Example (OpenAI SDK - Gemini through the gateway):
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ANTIGRAVITY_API_KEY"],
base_url="http://localhost:8045/v1",
)
resp = client.chat.completions.create(
model="gemini-2.5-pro",
messages=[{"role": "user", "content": "Hello"}],
)
Connecting Agent Clients¶
Claude Code - point the Anthropic env vars at the gateway:
export ANTHROPIC_API_KEY=$ANTIGRAVITY_API_KEY
export ANTHROPIC_BASE_URL=http://localhost:8045
claude
Any OpenAI-compatible client (Cherry Studio, Cursor, Continue.dev, or an OpenAI-compatible provider in an agent config): base URL http://localhost:8045/v1 plus the gateway API key.
CorpusIQ Use Cases¶
- Agent fleet failover - point OpenAI-compatible agents at the gateway for automatic 429/401 rotation across accounts.
- Model-cost arbitrage - remap hard-coded model names to cheaper tiers (
gpt-4.*→gemini-2.5-flash) without touching client code. - Quota monitoring - one dashboard for every Gemini/Claude account an agent fleet consumes.
- Premium-quota protection - background tasks auto-demote to Flash-tier models.
- Provider experimentation - test OpenAI/Anthropic/Gemini protocols against one local gateway before committing to a provider.
Security & Compliance Notes¶
- No skills.sh security audits published for this skill (no Trust Hub / Socket / Snyk verdicts render on the publisher page) - review before production use.
- The upstream tool repo (lbjlaq/Antigravity-Manager, 31K+ stars, pushed Sep 12 2026) ships no standard license (GitHub API: NOASSERTION).
- The gateway proxies web session tokens, not official API keys. Upstream providers may flag or ban accounts used this way (the built-in 403 detection exists for exactly this case). Review Google/Anthropic terms before production use.
Troubleshooting¶
| Issue | Fix |
|---|---|
| Port 8045 in use | lsof -ti:8045 | xargs kill -9 or change proxy.port in ~/.antigravity_tools/gui_config.json |
| Forgotten credentials | docker logs antigravity-manager or grep gui_config.json for api_key / admin_password |
| Account 403 / banned | Marked in UI and skipped; add a fresh account and remove the banned one |
| All accounts 429 | Rotation exhausted - add accounts or wait for quota reset; check the dashboard |
| macOS Gatekeeper | xattr -d com.apple.quarantine /Applications/Antigravity\ Tools.app |
| Claude Code not connecting | Verify the proxy responds and env vars are exported |
Verification¶
# Gateway responds on the models endpoint
curl -s http://localhost:8045/v1/models -H "Authorization: Bearer ${ANTIGRAVITY_API_KEY}"
# Full round-trip through the OpenAI-compatible endpoint
curl -s http://localhost:8045/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer ${ANTIGRAVITY_API_KEY}" \
-d '{"model":"gemini-2.5-pro","messages":[{"role":"user","content":"ping"}]}'
Notes¶
- From the ara.so Daily 2026 Skills collection - auto-generated from trending GitHub repos.
- Sibling guides from the same publisher: OpenClaw Control Center and Hermes Labyrinth Observability.
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.