Spend Control for Mastra
Mastra is the TypeScript agent framework for building stateful, tool-using agents — and tools are how agents spend. A guard keeps the tools honest.
Why Mastra agents overspend
Tool-calling agents hit paid endpoints as part of their loop.
Workflow steps can trigger purchases without a budget check.
Retries and reflection loops multiply tool calls.
How it works
Register the guard as a tool (or call the HTTP API from a custom tool). Before any spend, the agent asks sipi.bot: APPROVED, BLOCKED, or FLAGGED.
Rules that fit Mastra workloads
Per-transaction cap on tool purchases.
Merchant allowlist for known paid tools.
Velocity limit so loops die fast.
Guard tool
import { tool } from "mastra/core/tool";
const guardSpend = tool({
id: "guardSpend",
inputSchema: { amount: "number", merchant: "string", category: "string" },
execute: async ({ amount, merchant, category }) => {
const res = await fetch("https://sipi.bot/v1/transactions/evaluate", {
method: "POST", headers: { "Content-Type": "application/json" },
body: JSON.stringify({ amount, merchant, category }),
});
return (await res.json()).decision; // APPROVED | BLOCKED | FLAGGED
},
});One tool, deterministic policy.
FAQ
Does this work in TS runtimes?
Yes — it's a plain fetch to an HTTPS endpoint.
Can I cap a whole workflow?
Yes — per-agent rules apply to everything the agent spends.
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