How to Add Spending Limits to Your AI Agent in 5 Minutes

Prerequisites

1. Create an account at app.quetra.dev
2. Create an organization (happens automatically after signup)

That is it. No credit card required — the free tier includes 500 evaluations per month.

Path 1: MCP Server (Zero Code)

If your agent runs on Claude Desktop, Claude Code, or any MCP-compatible platform, this is the fastest path. No npm packages, no code changes.

Step 1: Register an Agent

Go to Agents → Register Agent in the dashboard. Give it a name and description. The description is critical — it is used by the MCP server for LLM agent routing when you have multiple agents.

Step 2: Create a Mandate

Go to Mandates → Create Mandate. Select your agent and configure the rules:

Click Activate to make it live. The mandate is now enforced on every transaction.

Step 3: Generate an API Key

Go to API Keys → Create Key. Copy the key (starts with sk_).

Step 3.5: Connect Your Stripe Account (Optional)

If your agent needs to make credit card charges via quetra_stripe_charge, go to Settings → Payment Integrations and add your Stripe secret key (sk_test_... for testing or sk_live_... for production). Your key is encrypted at rest and never exposed in API responses. Without this step, Stripe charges will run in mock mode.

Step 4: Connect

Your MCP server URL is:

https://mcp.quetra.dev/<your-api-key>/mcp

Add it to your platform:

• Claude Code: Settings → MCP Servers → Add Remote Server → paste the URL
• Claude Desktop: Settings → Connectors → Add → paste the URL

Done. Your agent now has 8 governance tools available: quetra_evaluate, quetra_check_budget, quetra_can_spend, quetra_fetch, quetra_transactions, quetra_list_agents, quetra_acp_checkout, and quetra_stripe_charge. The agent will check its budget before every purchase.

Path 2: SDK (Custom Agents)

For LangChain, CrewAI, custom frameworks, or any TypeScript/JavaScript agent, use the SDK.

Step 1: Install

npm install @quetra/sdk

Step 2: Set Up (Same as Above)

Register an agent, create a mandate with your rules, and generate an API key in the dashboard. Same steps as Path 1.

Step 3: Initialize the Client

import { QuetraClient } from '@quetra/sdk';

const quetra = new QuetraClient({
  apiKey: process.env.QUETRA_API_KEY,
  agentId: process.env.QUETRA_AGENT_ID,
});

Step 4: Check Before You Spend

// Pre-flight check — will this transaction be approved?
const canSpend = await quetra.canSpend({
  amount: 500,       // $5.00 in cents
  vendor: 'api.openai.com',
  category: 'generation',
});

if (canSpend) {
  // Proceed with the API call
  const result = await callOpenAI(prompt);
}

// Or use quetra.fetch() for automatic x402 payment handling
const data = await quetra.fetch('https://paid-api.example.com/research', {
  amount: 300,       // $3.00
  category: 'research',
});

Step 5: Check Budget Status

const budget = await quetra.getBudgetStatus();
console.log(budget);
// {
//   totalBudget: 10000,   // $100.00
//   spent: 2500,          // $25.00
//   remaining: 7500,      // $75.00
//   resetPeriod: 'monthly',
//   nextReset: '2026-05-01T00:00:00Z'
// }

What Happens When a Rule Fails?

The evaluation response tells you exactly which rule rejected the transaction and why:

{
  "approved": false,
  "reason": "vendor_blocklist",
  "details": "Vendor 'sketchy-api.com' is not in the allowlist",
  "rules": [
    { "type": "budget", "passed": true },
    { "type": "vendor_allowlist", "passed": false,
      "reason": "Vendor not in allowlist" },
    { "type": "rate_limit", "passed": true }
  ]
}

Every evaluation — approved or rejected — is logged to the transaction history with full context. Check it anytime via the dashboard or quetra.getTransactions().

Available Rule Types

Mandates support 7 rule types, combined with AND logic (all must pass):

Next Steps

• Quickstart guide — detailed setup with screenshots
• SDK reference — full API documentation
• Webhook integration — get notified on budget warnings and exhaustion
• What is AI agent spending governance? — the concepts behind the system