🔌 Claude Code + Cursor on your Google quota — $0, the bridge is open source

[SRZ]

THE AI CODING BILL → $0.00

 

One Google login, four top models, and your coding tools stay exactly as they are. The adapter is free, open source, MIT — and it landed on GitHub two days ago.

 

quota in → adapter → tools out — one glance, the whole trick

The tank came from this morning’s thread — the engine arrives here. One plan, both ends.

The trick in 10 seconds

• your Google AI Pro login already carries model quota — Google Antigravity is where it lives
• github.com/sksarvesh007/antiglaude turns that quota into a local API at 127.0.0.1:8080
• Claude Code · Cursor · LangChain · any OpenAI app plug straight in
• the tools you love stay; the subscription line on the bill goes

The model shelf, on one meter

you pick	what actually runs	born to
claude-sonnet-4-6	Claude Sonnet 4.6 · thinking	the daily coding agent
claude-opus-4-6-thinking	Claude Opus 4.6 · thinking	architecture, deep reasoning
gemini-3-flash	Gemini 3 Flash	speed runs, quick edits
gemini-2.5-pro	Gemini 2.5 Pro	large-context analysis

Run it tonight — 4 moves

1. pip install antiglaude — zero-install works too: uvx antiglaude start · pypi.org/project/antiglaude
2. sign in to antigravity.google once — the adapter lifts the login and renews its tokens forever
3. antiglaude start — your meter goes live at 127.0.0.1:8080
4. antiglaude setup-claude → open Claude Code → /model — the whole shelf sits in the picker

Cursor crew: add a custom OpenAI provider → http://127.0.0.1:8080/v1 — same shelf, your editor untouched.

the real panel: daily quota · latency log · request journal · keys r refresh · m switch model · c run Claude

• the meter reads your plan’s true ceiling — Pro tier shows ~1,500 requests a day — you always know how much fire you hold
• the pro move: a spare Google account carries the flag — the bridge is unofficial (Google’s terms name this OAuth route out of bounds), your main account stays a spectator

🧩 Hook up everything else

• Cursor / Continue.dev → custom OpenAI provider → http://127.0.0.1:8080/v1
• LangChain · LlamaIndex · AutoGen · CrewAI → base_url="http://127.0.0.1:8080/v1" + api_key="test" (the key is a placeholder — everything runs on your machine)

from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:8080/v1", api_key="test")
r = client.chat.completions.create(model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "ship it"}])
print(r.choices[0].message.content)

• want it alive before your IDE opens? antiglaude daemon install (macOS boots it for you)

📏 The fine print, read once

• the quota stays your plan’s quota — the dashboard counts the real number and invents nothing
• privacy by design: the gateway binds to loopback, traffic flows machine → Google, zero third-party telemetry
• long thinking turns survive — heartbeats keep the connection alive while Opus reasons
• antiglaude tui opens the full dashboard: quota meter, routing matrix, live journal
• on Windows and Linux antiglaude start is the whole story; the boot-on-login daemon is macOS’s extra

At 1Hack: the tank landed this morning, the engine arrived tonight — your quota, your tools, the paywall never enters the room.