Tutorial

How to Track Agent API Spending with Budgets

Build a budget tracker for AI agent API spending. Set limits per agent, alert on overruns, and track cost per task.

AI agents can burn through API budgets fast when left unsupervised. This budget tracker wraps your search API calls with cost tracking, sets per-agent spending limits, alerts before budget overruns, and logs cost-per-task for optimization. It works with any API that charges per call.

Prerequisites

  • Python 3.8+
  • requests library
  • A Scavio API key from scavio.dev
  • At least one agent making API calls

Walkthrough

Step 1: Build the budget-tracked API wrapper

Wrap API calls with cost tracking and budget enforcement.

Python
import os, requests, json, time
from datetime import datetime
from collections import defaultdict

API_KEY = os.environ['SCAVIO_API_KEY']
SH = {'x-api-key': API_KEY, 'Content-Type': 'application/json'}

class BudgetTracker:
    def __init__(self, cost_per_call=0.005):
        self.cost_per_call = cost_per_call
        self.agents = {}  # agent_id -> {budget, spent, calls}
        self.log = []

    def register_agent(self, agent_id, daily_budget):
        self.agents[agent_id] = {
            'budget': daily_budget,
            'spent': 0.0,
            'calls': 0,
            'created': datetime.now().strftime('%Y-%m-%d')
        }
        print(f'Registered: {agent_id} (budget: ${daily_budget}/day)')

    def can_spend(self, agent_id):
        agent = self.agents.get(agent_id)
        if not agent:
            return False
        remaining = agent['budget'] - agent['spent']
        return remaining >= self.cost_per_call

    def search(self, agent_id, query, **kwargs):
        if not self.can_spend(agent_id):
            remaining = self.agents[agent_id]['budget'] - self.agents[agent_id]['spent']
            print(f'BUDGET EXCEEDED: {agent_id} (${remaining:.3f} remaining)')
            return {'error': 'budget_exceeded', 'remaining': remaining}
        data = requests.post('https://api.scavio.dev/api/v1/search',
            headers=SH, json={'query': query, 'country_code': 'us', **kwargs}).json()
        self.agents[agent_id]['spent'] += self.cost_per_call
        self.agents[agent_id]['calls'] += 1
        self.log.append({'agent': agent_id, 'query': query, 'cost': self.cost_per_call,
                         'time': datetime.now().isoformat()})
        return data

tracker = BudgetTracker()
tracker.register_agent('research-agent', daily_budget=0.50)
tracker.register_agent('seo-agent', daily_budget=1.00)
tracker.register_agent('lead-agent', daily_budget=0.25)

Step 2: Use tracked search in agent workflows

Replace direct API calls with budget-tracked versions.

Python
# Simulate agent usage
queries = {
    'research-agent': ['best python framework 2026', 'fastapi vs django', 'python testing tools'],
    'seo-agent': ['serp api comparison', 'search api pricing', 'mcp search tool', 'ai agent search'],
    'lead-agent': ['saas companies hiring 2026', 'startup funding round'],
}

for agent_id, agent_queries in queries.items():
    print(f'\n--- {agent_id} ---')
    for q in agent_queries:
        result = tracker.search(agent_id, q)
        if 'error' in result:
            print(f'  BLOCKED: {q}')
        else:
            count = len(result.get('organic_results', []))
            print(f'  "{q}" -> {count} results')

# Check remaining budgets
print(f'\n=== Budget Status ===')
for agent_id, data in tracker.agents.items():
    remaining = data['budget'] - data['spent']
    pct = data['spent'] / data['budget'] * 100
    bar = '#' * int(pct / 5)
    print(f'  {agent_id:20} | ${data["spent"]:.3f}/${data["budget"]:.2f} ({pct:.0f}%) | {bar}')

Step 3: Generate spending reports

Analyze spending patterns and optimize agent budgets.

Python
def spending_report(tracker):
    print(f'\n=== Agent Spending Report - {datetime.now().strftime("%Y-%m-%d")} ===')
    total_spent = sum(a['spent'] for a in tracker.agents.values())
    total_budget = sum(a['budget'] for a in tracker.agents.values())
    total_calls = sum(a['calls'] for a in tracker.agents.values())
    print(f'  Total spent: ${total_spent:.3f} / ${total_budget:.2f} budget')
    print(f'  Total calls: {total_calls}')
    print(f'  Avg cost/call: ${total_spent/total_calls:.4f}' if total_calls else '')
    print(f'\n  Per-Agent Breakdown:')
    for agent_id, data in sorted(tracker.agents.items(), key=lambda x: x[1]['spent'], reverse=True):
        efficiency = data['spent'] / data['calls'] if data['calls'] else 0
        print(f'    {agent_id:20} | {data["calls"]:3} calls | ${data["spent"]:.3f} spent | ${efficiency:.4f}/call')
    # Monthly projection
    monthly = total_spent * 30
    print(f'\n  Monthly projection: ${monthly:.2f}')
    print(f'  Annual projection: ${monthly * 12:.2f}')
    # Optimization suggestions
    print(f'\n  Optimization Suggestions:')
    for agent_id, data in tracker.agents.items():
        usage = data['spent'] / data['budget'] * 100
        if usage < 30:
            print(f'    {agent_id}: Budget underused ({usage:.0f}%). Consider reducing to ${data["spent"] * 2:.2f}/day.')
        elif usage > 90:
            print(f'    {agent_id}: Near limit ({usage:.0f}%). Consider increasing budget.')

spending_report(tracker)

Python Example

Python
import os, requests
SH = {'x-api-key': os.environ['SCAVIO_API_KEY'], 'Content-Type': 'application/json'}

class Budget:
    def __init__(self, limit): self.limit = limit; self.spent = 0
    def search(self, q):
        if self.spent + 0.005 > self.limit:
            return print(f'Budget exceeded: ${self.spent:.3f}/${self.limit}')
        data = requests.post('https://api.scavio.dev/api/v1/search',
            headers=SH, json={'query': q, 'country_code': 'us'}).json()
        self.spent += 0.005
        print(f'{q}: {len(data.get("organic_results", []))} results (${self.spent:.3f}/${self.limit})')

b = Budget(0.025)
for q in ['api 1', 'api 2', 'api 3', 'api 4', 'api 5', 'api 6']:
    b.search(q)

JavaScript Example

JavaScript
const SH = { 'x-api-key': process.env.SCAVIO_API_KEY, 'Content-Type': 'application/json' };
let spent = 0; const limit = 0.025;
async function budgetSearch(q) {
  if (spent + 0.005 > limit) { console.log(`Budget exceeded`); return; }
  const data = await fetch('https://api.scavio.dev/api/v1/search', {
    method: 'POST', headers: SH, body: JSON.stringify({ query: q, country_code: 'us' })
  }).then(r => r.json());
  spent += 0.005;
  console.log(`${q}: ${(data.organic_results||[]).length} results ($${spent.toFixed(3)}/$${limit})`);
}
for (const q of ['test 1', 'test 2', 'test 3']) await budgetSearch(q);

Expected Output

JSON
Registered: research-agent (budget: $0.5/day)
Registered: seo-agent (budget: $1.0/day)
Registered: lead-agent (budget: $0.25/day)

--- research-agent ---
  "best python framework 2026" -> 10 results
  "fastapi vs django" -> 10 results

=== Budget Status ===
  research-agent       | $0.015/$0.50 (3%) | 
  seo-agent            | $0.020/$1.00 (2%) | 
  lead-agent           | $0.010/$0.25 (4%) | 

=== Agent Spending Report ===
  Total spent: $0.045 / $1.75 budget
  Total calls: 9
  Monthly projection: $1.35

Related Tutorials

Frequently Asked Questions

Most developers complete this tutorial in 15 to 30 minutes. You will need a Scavio API key (free tier works) and a working Python or JavaScript environment.

Python 3.8+. requests library. A Scavio API key from scavio.dev. At least one agent making API calls. A Scavio API key gives you 250 free credits per month.

Yes. The free tier includes 250 credits per month, which is more than enough to complete this tutorial and prototype a working solution.

Scavio has a native LangChain package (langchain-scavio), an MCP server, and a plain REST API that works with any HTTP client. This tutorial uses the raw REST API, but you can adapt to your framework of choice.

Start Building

Build a budget tracker for AI agent API spending. Set limits per agent, alert on overruns, and track cost per task.