Spend Control for the Google GenAI SDK
The GenAI SDK is the official Python/Go path to Gemini — and Gemini agents call tools that spend. The guard makes the spend policy real.
Why GenAI SDK agents overspend
Tool-calling agents hit paid endpoints per turn.
Long agentic sessions multiply inference and tool costs.
Gemini's own limits cover Gemini; the rest is ungoverned.
How it works
Register the guard as a tool (or call the HTTP API). Before any spend, the agent gets APPROVED, BLOCKED, or FLAGGED.
Rules that fit GenAI SDK workloads
Per-agent daily ceiling.
Merchant allowlist for paid tools.
Velocity limit for retry loops.
Guard call
from google import genai
from sipi_guard import sipi_guard
# Before a paid tool call:
decision = sipi_guard(amount=40, merchant="data-api.com", category="data")Gemini for the model; the firewall for the money.
FAQ
Does this replace Gemini's usage limits?
No — keep them for Gemini. The firewall covers every merchant, including outside Google.
Is there a Go client?
sipi.bot is HTTP — any language works.
Related
Stop the next $12,400 night.
One API call (or MCP tool) in front of every agent transaction — APPROVED, BLOCKED, or FLAGGED, deterministic, ~5 ms, fully logged.
See plans — from $99/mo Try a live check